فصلنامه پازند
پیاپی 30 (پاییز 1391)

Today, thanks to the systematic studies on the several linguistic corpora, most of the sociolinguists and discourse analysts believe that there are remarkable differences in the style and use of the language by men and women. However, such differences have not yet been analyzed thoroughly and statistically on the fictional prose of male and female writers. In the present article, computational approaches are employed to reach a stylistic objective on the gender-based differences between the use of language in several important novels written in English. First, a small corpus of some major English novels was formed. Then this corpus was analyzed on the basis of NLP from simple counting of the linguistic units, up to the more complicated statistical methods. The results of this research reveal that from the issues cited for the features of male and female's writings, female writers have made more use of the words about family, negative particles, and the preposition “for”. It I hoped that this article paves the way for similar studies on the Persian fiction analysis through corpus linguistics methods.

Parallel corpora regard as rich linguistic resources for Natural Language Processing and Cross Language Information Retrieval tasks. It is usually needed to align sentences before using these valuable resources; however, sentence alignments are expensive in terms of time and cost. With development of the World Wide Web and free access to it, automatically building parallel corpus from the Web is desirable. In this paper, we first choose bilingual pages with parallel content to extract parallel sentence candidates. Then, by computing several features and learning a Maximum Entropy classifier, parallel sentences are extracted from the candidate sentences. Our approach is not dependent on specific domain and it can cover different domains in the Web.

Spoken language understanding is considered as a specific domain of natural language understanding in which the uttered sentences are not as well-formed as written sentences. In the present paper, a text-based system of spoken language understanding is introduced for ticket reservation domain. This system is developed according to the datadriven approach and its architecture includes two main parts: first, extracting parameters of the model and second, assigning the most likely semantic tags to the sequence of words. "Hidden Markov Model" and "Viterbi" algorithm are applied in order to train the parameters and to tag the sequence of words. For this purpose, a corpus of commonly-used sentences in ticket reservation domain is collected and a specific tag is assigned to each word or a combination of words. In the training step, by using the tagged corpus, a sequence of possible tags is learned for a sequence of various words and in the testing step the most likely tag is assigned to a word or a combination of words according to the probabilities calculated in the previous step. Evaluation of the accuracy of system in recognizing the three key tags of departure, arrival and date is 91%.

Extracting semantic relations between the verb and its arguments in a sentence is useful for many natural language processing applications. On the other hand the selection restrictions which a verb applies on its arguments can be used in semantic processing of texts. Manual extraction of the argument structure of verbs besides the selection restrictions of all arguments of all verbs is very time consuming as well as costly. Thus automation of this task is one of the interests of researchers in semantic text processing field. In this paper, we propose three approaches to extract semantic relations between the verb and its arguments in a sentence. The first and the simplest one is based on the morphology and the lexical analysis of words. The second approach is based on generalization and extracts the selection restrictions by statistical study of the arguments. The third approach is based on rules and generalization which labels the semantic roles besides extracting the selection restrictions. After explaining the approaches, we compare them and discuss their pros and cons.

This paper introduces a system named “Sayeh-nama”. The overall idea is that such system is implemented for the first time in the field of metaphor for Persian. Sayeh-nama, with the purpose of helping the automatic understanding of metaphors, finds the shared semantic features of two terms in the metaphorical contexts. The overall process of this system consists of two phases. Firstly, the shared semantic features are extracted based on the second-order co-occurrence point-wise mutual information and then a number of adjectives are offered to describe these features. This method improves the performance of the system significantly in the systems of texts understanding, especially automatic understanding of the metaphors.

Church et al. (1991:12-13) have introduced some statistical measures such as Mutual Information (MI) test and t-score to identify significant lexical relations, especially to estimate associations between two words. On the other hand, Lyons (1995:62) has mentioned that the lack of absolute synonymy between the synonymous pair ‘big’ and ‘large’ is caused by the difference in their collocates. This paper intends to study the differences between synonymous pairs in terms of collocations using these two mentioned tests whilst mentioning the significance of linguistic corpora and the importance of statistical measures on linguistic studies.

There are differences between formal and colloquial varieties of languages in all aspects. The degree of the differences between formal and colloquial varieties is not similar in all languages. Persian is one of those languages in which the differences between formal and colloquial varieties are remarkable. This study investigates the phonological processes which turn the segmental string of formal Persian into colloquial one using the telephonic speech database, T-Farsdat in which the continuous speech has been segmented and annotated in two phonemic and phonetic levels. The alignment of these two strings gives us the type of phonological processes active in changing formal into colloquial Persian. Levenshtein Algorithm was used in aligning the phonemic and phonetic strings in order to show the type and frequency of the phonetic differences in formal and colloquial varieties of the language. The results of this study can be used in different aspects of theoretical study of the language and also in developing technologies for the language.