Paieškos sistemos „Google“ naudojimo galimybės tiriant kalbos reiškinius

SKAISTĖ ALEKSANDRAVIČIŪTĖ Institute of the Lithuanian Language POSSIBILITIES OF USING THE SEARCH ENGINE GOOGLE TO STUDY LANGUAGE PHENOMENA Lithuanian linguists increasingly note in their articles that, when collecting certain data—from examples of usage to numerical frequency measures—they have used some internet search engine. The largest search engine, Google, is mentioned most often, but, unfortunately, it is not sufficiently adapted for studying languages. It is therefore rather strange that authors usually limit themselves to referring to Google as one of the sources, without specifying the search method or examining certain features of the search engine and language that may facilitate or hinder the search, or clarify or distort the results. The search method is important both when collecting usage examples and when studying the prevalence of various linguistic phenomena, for which numerical measures are important. This article discusses the principles for constructing a search string, taking into account features of the search engine and the Lithuanian language that determine the success of a search. The article draws on a study of the relative frequency of new loanwords (hereafter, NS) conducted in 2007–2008, whose primary aim was to establish thresholds for classifying NS as rare, moderately frequent, or common. The study was based on the assumption that the frequency benchmarks established in the research would later prove useful in assessing the usage prevalence of other NS as well. This could also be useful in language standardization; accordingly, the aim of this article is to describe how the search string was constructed. The material for the study—NS found in the press and documents in 2005–2006 and on the internet in 20071—comprised 201 items in total. The list is highly varied: it contains 33 verbs and 168 nominals, including 25 nouns with defective paradigms (20 singular-only and 5 plural-only nouns), and includes both English and other 1The list was compiled in accordance with the 2006–2015 Programme for the Standardization, Use, Development and Dissemination of the State Language, using the card file of the project “Study of the Usage of New Lithuanian Loanwords: 1998–2008,” conducted by the Language Culture Department of the Institute of the Lithuanian Language. 266 KALBOS KULTŪRA | 81 language-origin NS. It is worth noting that some words studied as NS are included in the Dictionary of International Words (TŽŽ 2001) and may be considered international words. The internet is an extremely heterogeneous and constantly changing space. In conducting the study, internet websites were not categorized by either topic or style (although the Google search engine makes this possible), because the aim was to form a general picture of the prevalence of NS in general usage. Due to the specific characteristics of internet discourse, this article speaks in somewhat general terms, emphasizes the conditional nature of the findings, devotes considerable attention to possible errors, and is guided in many instances by linguistic intuition. The first part of the article presents the main principles for constructing a search string, both general principles and those specific to the study described here. The second part indicates ways of narrowing the search field and explains why this is necessary. 1. MAIN PRINCIPLES FOR FORMULATING A SEARCH QUERY Usually, a Google search is performed by entering the lemma of the word of interest in the search box. This search method is partly sufficient for collecting usage examples, but it is likely that a considerable number may go unnoticed, making it even less suitable for studying word frequency, where numerical values are important. The problem arises because the Google search engine partly understands that an entered word may have another form and includes such forms in the results. However, first of all, the search algorithm is not entirely clear to an ordinary user, and it is unclear how many and which word forms will appear among the results. Second, the search engine does not recognize parts of speech. For instance, if an infinitive is entered, nouns with the same root also appear among the results. For example, if the word autorizuoti is entered, the system will also return texts containing the word autorizacija. For the latter two reasons, it is preferable to conduct a somewhat more complex search that is, however, more useful for language research. Various tests have shown that the most effective approach is to enter the entire inflectional paradigm of a word, or most of it. The search query is limited by the fact that the Google search box can contain at most 32 keywords. S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 267 It is advisable to enclose the inflected forms of a word in quotation marks, since only text marked in this way is understood as an exact quotation by the Google system. All inflected forms of a word are joined by the logical operator OR (English or). A search query constructed in this way looks as follows: „akreditavimas“ OR „akreditavimo“ OR „akreditavimui“ OR „akreditavimą“ OR „akreditavimų“ OR „akreditavime“ OR „akreditavimai“ OR „akreditavimų“ OR „akreditavimams“ OR „akreditavimus“ OR „akreditavimais“ OR „akreditavimuose“ The aforementioned study of the relative frequency of NS was conducted using the Google search engine filter “pages in Lithuanian,” and has consequently already attracted criticism. The criticism is based on the claim that approximately two thirds of the information—that is, roughly that proportion of usage examples—is lost because not all website owners specify the language in their site metadata. However, this is entirely sufficient for studies of relative word frequency. Moreover, without using this filter, it is extremely difficult to exclude English-language texts returned by the search engine, which interfere with the collection of usage examples. We would therefore argue that searching web pages in Lithuanian is more appropriate; the important thing is to apply this restriction consistently throughout. Punctuation marks The Google search engine responds identically to any punctuation mark, even when it is placed in quotation marks and should be understood as an exact quotation. Thus, “account'as” entered in the search box will be treated as two separate words and conflated with “account:as,” “account as,” or other similar cases. This feature of the search engine matters in two situations. First, when an NS consists of two words, the system does not allow spelling variants such as “ex post” and “ex-post” to be distinguished; this could be important when assessing the use of the hyphen. In these cases, the frequency results are conflated, but when comparing the overall frequency of words rather than that of their individual spelling variants, it is not necessary to distinguish these forms in the search. Second, when spelling variants of NS with an apostrophe are studied, somewhat more complex problems arise, because some Lithuanian endings coincide with: a) certain English words (e.g., as, “like”; is, “is”; us, “us”); b) certain Lithuanian words (the conjunction o, the preposition į, and the interjections ei, ai); c) Lithuanian words written without diacritical marks (“shveplai,” i.e., in plain ASCII; is, “from”; as, “I”). 268 KALBOS KULTŪRA | 81 When conducting a search, it is very difficult to control these inaccuracies, while orthographic homography between NS endings and Lithuanian words is simply impossible to prevent. Nevertheless, cases in which one of the aforementioned Lithuanian words, whose written form coincides with the written form of an inflectional formant, immediately follows an NS are not so frequent as to have a significant effect on the research results. Selecting inflected forms for the search string Nouns and adjectives When studying the frequency of nominal NSs or their spelling variants, in most cases the search string included the complete inflectional paradigm for case and number. Exceptions were words with defective inflectional paradigms (20 singular-only and 5 plural-only nouns) and cases of polysemy (discussed in greater detail in the second part of the article). The vocative was also not worth including separately because of its orthographic overlap with other inflected forms and its low frequency. Thus, in this way, almost the entire inflectional paradigm of nouns and adjectives can be covered. When discussing the inflection of nominal NSs, the problematic issue of gender also needs to be addressed. In some cases, the feminine gender could be regarded as a spelling variant of an NS, for example, googlė / gūglė / google‘ė / googl‘ė; puzlė / puzzlė / pūzlė; skypė / skaipė. These NSs have in common that they denote inanimate objects and therefore do not inflect for gender in Lithuanian, and that their original forms end in –e, which is not pronounced but is represented in writing; cf. google, puzzle, skype. The study found that feminine forms in such cases are considerably less frequent than masculine forms; this can be explained by the assumption of optimality theory that “when there is a conflict, the least marked category is chosen” (Vaicekauskienė 2007: 177). The masculine and feminine forms of adjectives and nouns denoting animate objects were likewise not treated as separate lemmas, although there is some difference in meaning. Loreta Vaicekauskienė, who has studied NSs in Lithuanian in detail, notes that “[w]hen a noun can be used to name persons of both sexes, it seems that the masculine form is first generalized, from which a feminine form is sometimes created when needed” (Vaicekauskienė 2007: 175). Masculine and feminine forms were also not treated as separate lemmas when compiling Frequency Dictionary of Contemporary S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 269 rašomosios lietuvių kalbos žodyną (DDRLKŽ 1998: XIV–XV, XVI). Sudarant paieškos eilutes reikia į tai atsižvelgti, nes dirbant su paieškos sistema Google neįmanoma atskirti sutampančių vyriškosios ir moteriškosios giminės kait ybos formų. Verbs Verbs are a more complex case, because the number of grammatical categories of a verb is much greater. Since, as mentioned earlier, the length of a Google search string is limited, we need to ask which verb forms are the most representative in terms of frequency and should be included in the search string. It goes without saying that, to cover a wider distribution of verbs, the most frequently used forms need to be included in the search string. This requires data on the frequency of grammatical categories in Lithuanian, preferably in electronic discourse (ED). To date, ED as a distinct language variety alongside written and spoken language has been examined in greater detail by Asta Ryklienė, but her research methodology does not provide the required data on the distribution of grammatical categories in ED. The aim of her study was to compare ED with written and spoken language (Ryklienė 2001: 3), while quantitative differences in parts of speech and word forms between written language and ED were examined by comparing lists of the fifty words and word forms most frequently used in these language varieties (Ryklienė 2001: 9). This article draws on studies of word-form usage in Lithuanian by Vida Žilinskienė and Erika Rimkutė. However, their studies differ in that, unlike E. Rimkutė, V. Žilinskienė compares two functional styles—journalistic and technical—in her study of word-form frequency. On the assumption that the language of ED is closer to the journalistic than to the technical style, the author of this article draws on frequency statistics for grammatical forms in the journalistic style. The data reported by V. Žilinskienė and E. Rimkutė are very similar: the indicative mood accounts for 92.12% (Žilinskienė 2002: 110) and 91.5% (Rimkutė 2006: 35) of the overall verb distribution; finite verb forms account for 55.48% (Žilinskienė 2002: 110) and 50.7% (Rimkutė 2006: 35); and the infinitive accounts for 15.85% (Žilinskienė 2002: 111). However, E. Rimkutė (2006: 35) reports a somewhat different figure for the frequency of the infinitive: according to her, the infinitive accounts for 33.6% of non-finite verb forms, whose overall frequency is 49.3%. 270 KALBOS KULTŪRA | 81 In light of the latest data, the most convenient approach is to include in the search string all finite forms of the verb in the indicative mood, the infinitive, and the masculine plural nominative of the passive participle (compare the orthographic overlap with the infinitive). All 26 grammatical forms (21 distinct search terms in total, since the third-person forms of all tenses coincide) make up slightly more than half of the verb’s entire distribution. An example search string: “baninti” OR “baninu” OR “banini” OR “banina” OR “baniname” OR “baninate” OR “baninau” OR “baninai” OR “banino” OR “baninome” OR “baninote” OR “baninsiu” OR “baninsi” OR “banins” OR “baninsime” OR “baninsite” OR “banindavau” OR “banindavai” OR “banindavo” OR “banindavome” OR “banindavote” As mentioned earlier, the search engine Google ignores punctuation marks and treats spelling variants with an apostrophe as two separate words. Since the length of a Google search string is limited, when investigating verb spelling variants with an apostrophe, forms of the past frequentative tense may be omitted from the search string, since according to the data of both V. Žilinskienė and E. Rimkutė, the past frequentative is the rarest of all verb tenses. Of course, if there is a need to compare the frequency of different spelling variants of the same word, a broader search could be conducted, that is, several search strings could be constructed. However, when comparing the frequency of words rather than their inflected forms, the method described is arguably sufficient. It is only important to be consistent throughout and apply the same search criteria to all verbs. From what has been said in this subsection, it is clear that a search can cover more or less the entire inflectional paradigm of nominal words, but only slightly more than half of the inflectional paradigm of verbs; therefore, a sound basis for comparison is lacking. Of course, it would be possible to calculate a certain coefficient for the ratio between the frequencies of verbs and nominal words, but this would require more extensive ED research. For now, comparing the frequencies of individual verbs and nominal words would be inappropriate. Thus, the different sets of inflected forms for nominal words and verbs make it more convenient to separate them into two groups in frequency studies. S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 271 Spelling variants When investigating usage through internet searches, the process is the reverse of the corpus method, since constructing a search string requires one to decide in advance what spelling variants each word may have. When searching for linguistic phenomena online, authentic spellings of English origin should be avoided, and any that happen to occur in search strings should be removed. During the study mentioned above, it was observed that without doing so it is almost impossible to narrow the search field enough to prevent English insertions or phrases that should be considered quotations from appearing among the results. For example, searches for the words baby or crazy yield a great many texts mentioning song titles: Baby one more time, Crazy, and so on. The case of funky is similar, since this NS denotes not only a musical style but also a once very popular Lithuanian music group. Nevertheless, these spelling variants must be taken into account when assessing the relative frequency of such words, since they are fairly common in usage. The spelling variants investigated are those that seem possible based on linguistic intuition. Inflectional endings that are morphologically possible are added to an orthographically adapted borrowed word. When an NS is not orthographically adapted, inflectional endings are added directly to the indeclinable NS form. More unusual are cases in which the indeclinable NS form ends in a vowel, since, when investigating NS spelling variants without an apostrophe, that vowel is dropped (it turned out that smileas, skypeas, googleas, and similar spelling variants are rare in usage). When investigating linguistic phenomena that have not yet been standardized, the question often arises of where the boundary lies between a spelling error and a spelling variant. This boundary is determined by frequency. For example, in the NS study mentioned above, NS spelling variants with fewer than 10 occurrences were excluded from the calculation of the overall frequency of each NS and considered spelling errors—isolated slips that do not reflect general usage tendencies. It is true that this principle was applied only to words exceeding 1,000 occurrences, and not to rarer words. Nevertheless, it can sometimes be significant that the linguistic phenomenon under investigation was not found online at all, or that it did not reach 10 occurrences. The numerical expressions denoting the frequency of the spelling variants constructed and selected according to these principles are summed to obtain an approximate word frequency. The following table presents the processing of search results more clearly. 272 KALBOS KULTŪRA | 81 TABLE 1. Data collection method Spelling variant of NS Search string for an individual spelling variant Frequency of the NS spelling variant puzlas „puzlas“ OR „puzlo“ OR „puzliui“ OR „puzlą“ OR „puzlų“ OR „puzle“ OR „puzlai“ OR „puzlų“ OR „puzlams“ OR „puzlus“ OR „puzlais“ OR „puzluose“ 6430 puzlė „puzlė“ OR „puzlės“ OR „puzlei“ OR „puzlę“ OR „puzlėje“ OR „puzlėms“ OR „puzles“ OR „puzlėmis“ OR „puzlėse“ 3210 puzlis „puzlis“ OR „puzlio“ OR „puzliui“ OR „puzlį“ OR „puzliu“ OR „puzlyje“ OR „puzliai“ OR „puzlių“ OR „puzliams“ OR „puzlius“ OR „puzliais“ OR „puzliuose“ 711 puzzlė „puzzlė OR „puzzlės“ OR „puzzlei“ OR „puzzlę“ OR „puzzlėje“ OR „puzzlėms“ OR „puzzlėmis“ OR „puzzlėse“ 458 puzzlis „puzzlis“ OR „puzzlio“ OR „puzzliui“ OR „puzzlį“ OR „puzzliu“ OR „puzzlyje“ OR „puzzliai“ OR „puzzlių“ OR „puzzliams“ OR „puzzlius“ OR „puzzliais“ OR „puzzliuose“ 315 pūzle „pūzle“ OR „pūzlės“ OR „pūzlei“ OR „pūzlę“ OR „pūzlėje“ OR „pūzlėms“ OR „pūzles“ OR „pūzlėmis“ OR „pūzlėse“ 239 puzzlas „puzzlas“ OR „puzzlo“ OR „puzzlui“ OR „puzzlą“ OR „puzz-lų“ OR „puzzlai“ OR „puzzlų“ OR „puzzlams“ OR „puzzlus“ OR „puzzlais“ OR „puzzluose“ 167 pūzlis „pūzlis“ OR „pūzlio“ OR „pūzliui“ OR „pūzlį“ OR „pūzliu“ OR „pūzlyje“ OR „pūzliai“ OR „pūzlių“ OR „pūzliams“ OR „pūzlius“ OR „pūzliais“ OR „pūzliuose“ 101 puzzle’as „puzzle’as“ OR „puzzle’o“ OR „puzzle’ui“ OR „puzzle’ą“ OR „puzzle’u“ OR „puzzle’ė“ OR „puzzle’ai“ OR „puzzle’ų“ OR „puzzle’ams“ OR „puzzle’us“ OR „puzzle’ais“ OR „puzzle’uose“ -are 80 11 711 A similar table was compiled for each investigated NS. The first column lists all the spelling variants of the investigated NS, ranked by frequency, since S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 273 the article refrains from considering which spelling variant of an NS could be regarded as the headword (lemma). The second column gives the search queries for each spelling variant after narrowing the search field, i.e. after removing all ambiguous forms (see Section 2). The third column indicates the frequency of each spelling variant (these data are only intermediate), which is later added up to obtain an approximate frequency for the word under investigation. Only the latter data are significant for determining the limits of relative NS frequency (the data in the table were collected in March–April 2008). 2. NARROWING THE SEARCH FIELD On the excess of results in English The problem of English homographs (words with identical spelling) and spelling variants containing an apostrophe can be partly solved by using the logical “except” function, which in the Google system is represented by the “-” sign. This means that the search will not be carried out on webpages containing the word following this sign. Therefore, by adding one of the most frequent English words at the end of the search query and marking it with the sign mentioned above, the number of English-language occurrences can be greatly reduced. In the aforementioned study, the exclusion of one of the most frequently used words in English, are2 (are – third-person plural present tense), was applied throughout, but another choice is also possible. It should merely be noted that an English word must be chosen very carefully for this purpose, since some of the most frequent English words may be orthographically identical to some Lithuanian words (English it, be), inflectional endings (English is, as, a, I), or “unaccented” spellings (English is, as, I). In addition, many of the most frequent English words (a, the, of) serve a grammatical function and are often used in various names, which may also occur on Lithuanian websites. The same restriction applies in all cases where the search query consists of an invariable NS form (with authentic or adapted spelling), regardless of the language from which the particular word may have entered Lithuanian. 2 The 500 Most Commonly Used Words in the English Language: http://www.world-english.org/english500.htm (accessed 2008 06 07). 274 KALBOS KULTŪRA | 81 It has been observed that even when searching for NS of non-English origin and only on Lithuanian-language websites, many examples in English still appear among the results. This is associated with the language’s lingua franca status in ED. Moreover, some of the NS examined may already have become international words in English or may simply be non-adaptable. On the polysemy of words and their forms When using the Google search engine, one inevitably encounters polysemous words or their forms. The term polysemy3 was chosen, first of all, to avoid confusion among terms such as homonyms, homographs, homophones, and the like in Lithuanian linguistics4. Second, explanations of the definitions of these terms often include the phonetic aspect of a word, which is entirely irrelevant when conducting searches on the internet. Third, the aforementioned terms are insufficient to designate certain cases of polysemy, for example, when inflectional forms of the same word, but belonging to different spelling variants, are identical in spelling. Polysemy may occur at several levels: derivational (the same derivational affix conveys different meanings), textual, syntactic-semantic, lexical-semantic (subdivided into homonyms, homographs, homophones, and polysemy) and morphological (Rimkutė 2006: 7, 40). This article is concerned with morphological polysemy (hereafter MD) and, in part, lexical-semantic polysemy. Morphologically polysemous forms of the same word Morphological polysemy among forms of the same word is a systematic phenomenon, as it is associated with the characteristics of the Lithuanian morphological system and case syncretism. This article lists only those cases of morphological polysemy encountered during the research and indicates how to handle them in the Google system. 3 E. Rimkutė also uses the term polysemy and discusses this phenomenon extensively in her works (Rimkutė 2002, 2003; Rimkutė, Grybinaitė 2004). 4 Cf. the definitions of homonyms, homographs, homophones, and homophones in these sources, as well as their usage characteristics: Jakaitienė 1980: 42–48; 2005: 53–55, 161–168; Gaivenis, Keinys 1990: 79–80; LKE 1999: 247–249. S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 275 Inflectional forms that are orthographically identical within the same spelling variant: Nouns with an (i)o stem in the singular nominative and instrumental, e.g.: accreditation, alibidemia, aromatherapy, authorization, blogosphere / blogsphere, digitalization, harissa / harisa, convergence, looserish / loserish, media / media, milonga, salsa, spammer (feminine); Nouns with an (i)o stem in the singular genitive and plural nominative, e.g.: accreditation, authorization, blogosphere / blogsphere, harissa / harissa, looserish / loserish, media / media, female spammers; Nouns with an ė stem in the singular genitive and plural nominative, e.g.: bikers (female), bloggers / blogger’s / blog’ers, glossy magazines, goalball players (female), karaoke, losers (female) / looseresses / looser’esses / loser’esses / loseresses, managers (female) / manageresses, multiplexes (female), nonconformists (female), puzzles (see Table 1), Skype accounts, sponsors (female), showwomen; Nouns with an (i)u stem in the singular nominative and plural genitive, e.g.: aspirator, blocker, epilator / epile tor; Nouns with an (i)u stem in the singular accusative and plural genitive, e.g.: aspirators, sponsors. These MD cases may also be called grammatical homonyms, which are defined in Lithuanian linguistics as “phonologically identical forms of the same word that have different grammatical meanings” (Jakaitienė 1980: 42). What these forms have in common is that, when a search is carried out in Google, they appear in the same search string, are joined by the logical operator OR, and their frequencies are not summed again. In such cases, it is still convenient to shorten the search string for efficiency, but this does not affect the results. Another type of MD consists of inflectional forms that are orthographically identical for the same word but belong to different spelling variants: plural genitives of masculine and feminine nouns, e.g.: bikers, bloggers / blogger’iu / blog’ers, glossy ones, losers / loosers / loosers / loser’iu / losers, looserish / loserish, managers / manageriu, multiplexes, puzzles / puzles / puzzless (see Table 1), spammers, sponsors; nouns with an (i)a stem in the singular nominative and nouns with an (i)o stem in the plural accusative, e.g.: looser iškas / lūzeriškas, spammer; 276 KALBOS KULTŪRA | 81 nouns of the (i)a stem in the plural nominative and nouns of the (i)o stem in the singular dative, e.g.: spammers; nouns of the (i)a stem in the singular locative and nouns of the ė stem in the singular instrumental, e.g.: gūglė / google‘ė, internaute, puzle / pūzle (see Table 1), skaipe, šo(u)mene; nouns of the (i)o and (i)a stems in the singular accusative, e.g.: a looser form / a loser-like form, a spammer. These latter MD cases are more complex and require greater care. Inflectional forms of different spelling variants of the same word that are orthographically identical occur in different search queries; this means that, when calculating the overall frequency of an individual word, the frequencies of such forms are counted more than once. Therefore, when studying word frequency, overlapping forms of this type should be retained in only one of the search queries. It is worth mentioning that, in conducting a study of the relative frequency of NS, two previously undescribed types of MD were encountered, in which the plural nominative of nouns of the (i)a stem is orthographically identical to the singular dative of nouns of the (i)o stem (cf. tie varnai and tai várnai), and the singular locative of nouns of the (i)a stem is orthographically identical to the singular instrumental of nouns of the ė stem (cf. tame draugė5 and su draugė). Although these two types of MD were identified when studying the use of NS, the examples given show that they are also characteristic of standard Lithuanian. Systematic MD cases that nevertheless fall outside the scope of Lithuanian are inflectional forms of morphologically adapted NS that are orthographically identical to uninflected, authentic-spelling NS forms not under investigation: nouns of the (i)a stem in the singular locative, e.g.: drive (drivas), google (googlas), puzzle (puzzlas, see Table 1), skype (skypas); nouns of the (i)o stem in the singular nominative and singular instrumental, e.g.: media; nouns of the ė stem in the singular instrumental, e.g.: google (googlie), karaoke (karaoke), puzzle (puzzle, see Table 1), skype (skype). Because, for the reasons mentioned earlier, NS variants in authentic spelling are difficult to investigate, morphologically ambiguous forms of the latter type must be removed from search strings in all cases. 5 Although this form is not to be used from the perspective of language culture, it is theoretically possible. S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 277 Morphologically ambiguous word forms with different lexical meanings Sometimes all or some inflected forms of one word may be spelled identically to inflected forms of another word. Clearly, unlike the MD cases described earlier, the lexical meaning will differ. Moreover, these MD cases are accidental and not systematic. The following examples of this type of MD found during the study are listed, indicating the NS spelling variant form and the standard-language word with which that form coincides: – blogas, b l o g a s (adjective), b l o g a i (adverb) – the search string is shortened to “blogui” OR “bloge” OR “blogams” (only forms differing from those of the standard-language word are listed – singular dative, singular locative, plural dative); – smailas, s m a i l a s (adjective), s m a i l a i (adverb) – the search string is shortened as in the case of blog: “smailui” OR “smaile” OR “smailams”; – bodis, b o d u s (adjective), b o d ž i a i (adverb) – the search string is shortened to “bodis” OR “bodžio” OR “bodžiui” OR “bodį” OR “bodyje” OR “bodžiams” (forms differing from those of the standard-language word – singular nominative, singular genitive, singular dative, singular accusative, singular locative, plural dative); – brendas, b r e n d o (bręsti, bręsta, brendo), b r e n d u (bristi, brenda, brido); – there were also cases in which the NS spelling variant under investigation coincided completely with a standard-language word – cf. bōtas, b ò t a s (a tall women’s galosh), b õ t a s (a type of ship), as well as sáitas and s a ì t a s. Data provided by the VDU corpus or DDRLKŽ on the distribution of words can help determine how to handle such cases of homonymy. If the coinciding standard-language word is rare, the search string need not be shortened, since this has little effect on the search results. Moreover, sometimes results for just one spelling variant are sufficient to regard the word under investigation as frequent. For other reasons There are cases when the word under investigation may be used in some widely used company, institution, or other name, or chosen as a pseudonym. The Internet is usually awash with names; this is more closely linked to advertising strategies or the structural principles of discussion forums than to the usage habits of the words used to form those names. Therefore, when conducting a search, efforts are made to exclude such cases. 278 KALBOS KULTŪRA | 81 The search field can be narrowed in two ways. First, by context, when the search query specifies that a word should be excluded if, according to usage examples, it is often used in the same context as the linguistic phenomenon under investigation but does not form a word combination with it. Second, by collocation6, when the search query specifies that a word should be excluded if it forms a frequently used word combination in actual usage with the phenomenon under investigation. The more unusual cases of narrowing the search field encountered in the study are presented in Table 2, listing the words used, as well as the spelling variants and forms of the NS to which the narrowing was applied (collocations are also indicated where necessary). TABLE 2. Narrowing the search field by context and collocation By context -nikas -author -user -member -blog -uab -office -plot Harissa / harisa, googlis, sasha funke, esesas, twisteris, neformatas, mixas, extremas topas, saitas, respectas, funkaś, autletas, bugo grafitas skypas By collocation -la -center -eesti -area -baltic +coffee –western „La speme“ (spemas) „Topo centras“ (topas) „Eesti post“ (post’as) „Cluster area“ (cluster‘is) „Baltijos autletas“ (autletas) Ekspreso kava, Vakarų ekspresas Most difficulties arise with words used as pseudonyms (the first and second rows of the table). Depending on the search results, one or more of the specified exclusion terms must be selected. The table lists only the most effective exclusion terms, but the final choice depends on the specifics of each case under investigation. 6 Collocation is a very typical context for a word’s use, meaningful and grammatical, and a combination of at least two words that are often used together (Jakaitienė 2005: 25). S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 279 In this context, the case of skaipo is also very interesting and somewhat unusual. When investigating the spelling variant skypas of this NS, it was necessary to narrow the search field because numerous online advertisements for plots of land for sale were found, with a misspelling (the letter l had been omitted). CONCLUSIONS 1. When constructing a search string, the selection of representative word forms is of primary importance; this is carried out on the basis of the results of studies on the frequency of grammatical categories in Lithuanian. When investigating nominal words, almost the entire inflectional paradigm fits into the search string, whereas for verbs only part of it does. The most practical approach is to enter the infinitive and the finite forms of the verb tenses in the indicative mood (present, simple past, future, and frequentative past) into the search string. This makes it possible to cover somewhat more than half of the distribution of verb grammatical categories. 2. Because the search string has a limited length, the coverage of grammatical-category distributions differs, depriving us of a sound basis for comparing verbs and nominal words. It is therefore more practical to separate the nominal-word and verb groups. Google 3. The search field often has to be narrowed because of the excess of English-language results and the problem of polysemy in Lithuanian. This is done using the logical “exclude” function, which is represented in the syntax of the Google search engine by the “-” symbol, or by omitting certain polysemous forms from the search string. 4. When conducting a search, attention should be paid to cases of polysemy that affect the search results in different ways and therefore call for different restrictions on the search field. Orthographically identical inflected forms of the same spelling variant can be removed from the search string without affecting the results. It is advisable to retain inflected forms of the same word that differ between spelling variants in the search string for only one of the variants. Morphologically polysemous forms of words with different lexical meanings should be retained or removed in light of data from the VDU corpus and DDRLKŽ. 5. The search method discussed is useful both for collecting examples of various linguistic phenomena and for investigating their prevalence in usage. In addition, the method for determining the relative frequency of NS can also be applied in language standardization, for example, when compiling lists of NS regarded as major language errors. 280 KALBOS KULTŪRA | 81 ŠALTINIAI Paieškos sistema www.google.lt The 500 Most Commonly Used Words in the English Language: http://www.world-english.org/english500.htm TŽŽ 2001 – Tarptautinių žodžių žodynas, Vilnius: Alma littera. VDU Dabartinės lietuvių kalbos tekstynas: http://donelaitis.vdu.lt LITERATŪRA DDRLKŽ 1998: L. Grumadienė, V. Žilinskienė. Dažninis dabartinės rašomosios lietuvių kalbos žodynas, Vilnius: Lietuvių kalbos institutas. Gaivenis K., Keinys S. 1990: Kalbotyros terminų žodynas, Kaunas: Šviesa. Jakaitienė E. 1980: Lietuvių kalbos leksikologija, Vilnius: Mokslas. Jakaitienė E. 2005: Leksikografija, Vilnius: Mokslo ir enciklopedijų leidybos institutas. LKE 1999: Lietuvių kalbos enciklopedija, Vilnius: Mokslo ir enciklopedijų leidybos institutas. Rimkutė E. 2002: Homoformos dabartinės lietuvių kalbos tekstyne. – Lituanistica 2(50), 86–101. Rimkutė E. 2003: Morfologinio daugiareikšmiškumo tipologija. – Lituanistica 4(56), 60–78. Rimkutė E. 2006: Morfologinio daugiareikšmiškumo ribojimas kompiuteriniame tekstyne. Daktaro disertacija, Kaunas. Rimkutė E., Grybinaitė A. 2004: Dažniausios lietuvių kalbos morfologinio daugiareikšmiškumo rūšys ir jų automatinis vienareikšminimas. – Kalbų studijos 5, 74–78. Ryklienė A. 2001: Elektroninis diskursas: kalbos ypatybės ir stilius. Daktaro disertacijos santrauka, Vilnius. Vaicekauskienė L. 2007: Naujieji lietuvių kalbos svetimžodžiai. Kalbos politika ir vartosena, Vilnius: Lietuvių kalbos institutas. Žilinskienė V. 2002: Žodžių formų vartojimas lietuvių kalbos dalykinio ir publicistinio stilių duomenimis. – Lituanistica 1(49), 106–116. Gauta 2008 07 08 S. ALEKSANDRAVIČIŪTĖ. Paieškos sistemos Google naudojimo galimybės tiriant kalbos reiškinius 281 POSSIBILITIES OF USING THE SEARCH SYSTEM GOOGLE IN LINGUISTIC RESEARCH S u m m a r y The paper introduces the methodology which can be used to implement Google as a tool for data accumulation and storage and investigating the spread of a linguistic phenomenon in the internet. The method is described in the framework of an investigation aimed at identifying relative frequency count of new loanwords between 2007 and 2008. The first section of the article describes the main principles of filling in the slot of a query: general as well as specific for a particular descriptive research. Rather than entering only the main word form, a proposal has been launched to enter the whole grammatical paradigm of the word or its major part considering that in the Google query slot the maximum number of key words is 32. The second section identifies the ways to narrow down the search domain and reasons why it should be done. This type of research is bound to face a problem of word and word-form ambiguity. The term of ambiguity has been selected to avoid confusion of more specific terms like homonyms, homographs, homophones etc. Moreover, the explication of the above terms often emphasises the phonetic aspect of the word, which for internet search is irrelevant. The above terms are not sufficiently precise for some cases of ambiguity, when, for example, the spelling of one word form coincides with another form of the same word. The method of search is instrumental in collecting various examples of language use as well as in investigating their frequency. Also the method of relative frequency of loanwords could be applied in language standardisation, e.g. in compiling lists of new loanwords which are considered grave mistakes. SKAISTĖ ALEKSANDRAVIČIŪTĖ Lietuvių kalbos institutas P. Vileišio g. 5, LT-10308 Vilnius [email protected] 282 KALBOS KULTŪRA | 81