SYNTACTIC PARSING

A stage in the processing of written or spoken language at which a syntactic structure is assembled from a string of words. Two problems arise in attempting to relate syntactic parsing to linguistic theory:

 a. Linguistic theory attempts to describe how language is structured, not how a language user performs. It does not claim that the way in which it chooses to represent a linguistic system is psychologically real (a model of how the user’s mind operates). Linguistics attempts to represent language competence, while psychology attempts to describe language performance.

b. Modern approaches to syntax are based upon hierarchies of phrases– reflecting the structure dependency that characterises all languages. But the psychological processes of reading and listening have been shown to be linear, with words decoded as they are perceived. A theory of parsing has somehow to square these two perspectives.

& ‘Linguistic’ approaches. Early approaches to parsing took transformational grammar as a point of departure and attempted to establish that Chomsky’s descriptive system represented what actually takes place in the mind of the language user. The primary role of parsing was seen as being to recover the deep structure of a sentence by reversing any transformations that might have gone into its creation. A Derivational Theory of Complexity (DTC) hypothesised that the more transformations there were, the more difficult it was for a listener or reader to process a sentence. A task in which subjects matched derived sentences to their deep structure form seemed to support the theory. However, there were many uncontrolled variables in the material. Later commentators dismissed the DTC theory on a number of counts: among them, the fact that listeners do not wait until the end of a sentence in order to begin processing it.

 Researchers next tried to confirm the psychological reality of deep structure. ‘Click experiments’ were devised in which listeners had to identify the location of a click inserted into sentences with differing deep structures. A click was inserted during the word criminals in:

 [The corrupt police can’t bear] [criminals to confess quickly]

 [The corrupt police can’t force criminals] to confess quickly

Subjects reported it as occurring before the word in sentence a. and afterwards in sentence b., suggesting that they had indeed constructed a representation based on deep structure. However, again there were problems of research design– not least the fact that subjects reported post-perceptually– a short time after hearing the click.

Much research into syntactic parsing still takes Chomskyan theory as its framework, and continues to seek for evidence of d structure (deep structure) patterns underlying s-structure (surface structure) ones. In current theory, the concept of transformation has been replaced by one of movement. A sentence such as:

I wonder which book my cousin borrowed (t)

is said to be the product of a movement rule which shifts the Noun Phrase which book leftwards from the position it originally occupied in deep structure. That position is marked in the example above by (t), showing where a trace of the NP remains. Researchers have suggested that the successful parsing of a sentence such as c. depends upon temporarily storing which book in memory until its deep structure position (t) is reached. Evidence from timed reading supports this theory: reading slows when a ‘filler’ such as which book is encountered, suggesting that demands on working memory do indeed increase at that point.

 Another area of research influenced by Chomskyan theory seeks evidence of the setting of parameters, linguistic features which are adjusted according to the language which one acquires. One line of enquiry compares syntactic competence across speakers of different languages, by asking them to make grammaticality judgements. There has, for example, been interest in the processing of wh- interrogatives involving prepositions since languages vary widely in the degree to which they permit preposition stranding (Who did he speak to?) as against pied piping (To whom did he speak?).

 Another line of enquiry attempts to establish when parameters become established in young language acquirers. There has been interest in anaphor resolution, particularly where it involves reflexive pronouns. At an early age, children have shown themselves much more successful in interpreting a sentence containing a reflexive (Cinderella’s sister points to herself) than a similar sentence containing an ordinary pronoun (Cinderella’s sister points to her).

The unit of parsing. Experiments investigating wrap up effects demonstrate that it becomes difficult for subjects to report the verbatim form of a spoken utterance once a clause boundary has been passed. The conclusion is that meaning is ‘packaged’ clause by clause and that the actual words heard are then lost to memory since theyare no longer required. There is also evidence from reading that the processing load on working memory is greatest at the end of a clause and that eye fixations are longer. These effects appear to reflect the reader’s need to allocate additional attention to syntax. It thus seems that the clause boundary is the point at which syntactic decisions are made.

Linear processing. But processing is not delayed until a clause is complete, as early parsing theory assumed. Indeed, experiments have established that both listening and reading take place on line, with the language user processing words as they are identified before going on to construct higher-level syntactic patterns from them. Parsing is therefore viewed as an incremental process.

 Researchers in Artificial Intelligence have addressed the issue of how to combine linear processing at word level with the building of higher-level syntactic structures. They have devised computer programs known as Augmented Transition Networks (ATNs) which identify a series of points in the parsing of an utterance at which accumulated evidence can lead to a ‘change of state’ and thus a new direction for the processing. The points occur after each complete phrase and sentence as well as after each word, thus enabling syntactic structures to be built. While computer networks of this kind enable programs to process pieces of text, they are heavily dependent upon conventional sentence structures, and their top-down nature leads to many incorrect predictions and much backtracking.

Ambiguity. Much parsing research has investigated how the language user deals with cases of ambiguity as typified by garden path sentences such as:

        d. The old man / the boats. (slash indicates point of ambiguity)

        e. John remembered the answer / was in the book

where the word class of a word and/or the syntactic structure of a sentence are unclear at a given point in the sentence but are subsequently resolved. Evidence suggests that a preferred analysis is chosen by the parser and later revised if necessary. There are several theories as to why a particular interpretation is preferred:

Syntactic. The listener/reader exercises a preference for a canonical (SVO) sentence structure. Alternatively, specific syntactic strategies are used, such as ‘make an attachment to the clause that is currently being processed’.

Lexical. The preferred reading reflects the argument structure of the current verb.

Semantic. World knowledge indicates which interpretation is the most probable.

Syntactic vs semantic processing. The discussion of ambiguity exemplifies a broader disagreement over the extent to which semantic criteria influence syntactic parsing. The major issue is whether parsing involves syntax and semantics together or whether a first pass analysis relies solely on syntax. Strong evidence for separate processes comes from studies of the brain. A negative event-related potential (variation in electrical activity) occurs 400 milliseconds after a semantic anomaly has been detected; but a very different positive ERPoccurs 600 milliseconds after a syntactic anomaly. Evidence from readers’ eye movements is also cited in support of a two-stage model. Subjects appear to be slowed down by sentences such as g below, where semantic evidence of animacy is available to help them, just as they are by f, where it is not.

f. The defendant examined by the lawyer was unreliable.

g. The evidence examined by the lawyer was unreliable.

See also: Ambiguity: syntactic, Anaphor resolution, Augmented Transition Network, Derivational theory of complexity, Event related potential, Garden path sentences, Wrap up effects

Further reading: Aitchison (1998: Chaps 9–10); Harley (2001: Chap. 9); Pickering (1999)

اخر الاخبار

اخبار العتبة العباسية المقدسة

الحزام الأخضر الجنوبي الأول يستضيف وفدًا شبابيًّا من محافظة المثنى

قسم شؤون المعارف الإسلامية يحتفي بإرث الأديب الراحل علي أنكيل ويتسلم مكتبته الشخصية

سماحة السيد الصافي يدعو إلى توثيق الأفكار والأحداث لحماية الإرث العراقي من الاندثار بغية نقله سليمًا للأجيال القادمة

قسم التطوير يقدم محاضرة ثقافية حول كتاب نهج البلاغة لملاكاته

المزيد

الآخبار الصحية

أول إرشادات غذائية قائمة على الأدلة لمعالجة الإمساك المزمن

صباح بلا إفطار؟.. صحتك في خطر!

هل يمكن للجلد أن يكشف عن صحتك العقلية؟.. دراسة حديثة تجيب

السر في مطبخك.. 6 أطعمة تمنح عظامك قوة الشباب

المزيد

الآخبار التكنلوجية

ناسا تكشف: كوكب الأرض أصبح أقل لمعانا خلال العقود الماضية

دراسة: الوقت الذي تستخدم فيه هاتفك قد يفضح حالتك النفسية

الهنغاري لاسلو كراسناهوركاي يفوز بجائزة نوبل للآداب 2025

رسميا.. إعلان الفائز بجائزة نوبل للسلام لعام 2025

المزيد

مواضيع ذات صلة

SYNTACTIC DEVELOPMENT

SYNAESTHESIA

SYMBOLIC REPRESENTATION

SW (STRONG-WEAK) PATTERN

SUBVOCALISATION

STUTTERING

STROOP TEST

STRATEGY

STORY GRAMMAR

SPREADING ACTIVATION

SPEECH: UNIT OF PRODUCTION

SPEECH SIGNAL