International Journal of Computer Processing of Languages

The lack of internal information of Chinese synthetic words has become a crucial problem for Chinese morphological analysis systems, which are facing various needs of segmentation standards for upper NLP applications being developed. In this paper, we first define the conceptual differences between Chinese single-morpheme words and Chinese synthetic words. Then we define Chinese synthetic words into two types, compound words and morphologically derived words, according to their internal syntactic and morphological structure and classify them into more specific categories. After making a survey on three-character Chinese synthetic words based on these categories, we propose a tree-based analysis method to represent the internal information of the words. Next, we use machine learning methods to automatically identify the internal morphological structure of three-character synthetic words by using a large corpus and add syntactic tags to their internal structure. We believe that the tree-based word internal information is useful in specifying a Chinese synthetic word segmentation standard. We also believe that the internal information of Chinese synthetic words can help to improve morphological analysis and out-of-vocabulary (OOV) word detection of Chinese text.

Since the first Chinese Word Segmentation (CWS) Bakeoff in 2003, CWS has experienced a prominent flourish because Bakeoff provides a platform for the participants, which helps them recognize the merits and drawbacks of their segmenters. However, the evaluation metric of Bakeoff is not sufficient to measure the performance thoroughly; sometimes it is even misleading. One typical example caused by this insufficiency is that there is a popular belief existing in the research field that segmentation based on word can yield a better result than character-based tagging (CT) on in-vocabulary (IV) word segmentation even within closed tests of Bakeoff. Many efforts were paid to balance the performance on IV and out-of-vocabulary (OOV) words by combining these two methods according to this belief. In this paper, we provide a more detailed evaluation metric of IV and OOV words than Bakeoff to analyze CT method and combination method, which is a typical way to seek such a balance. Our evaluation metric shows that CT outperforms dictionary-based (or so-called word-based in general) segmentation on both IV and OOV words within Bakeoff closed tests. Furthermore, our analysis shows that using confidence measure to combine the two segmentation results should be under certain limitation.

We present a machine learning approach for coreference resolution of noun phrases. In our method, we use CRFs as a basic training model, and use active learning method to generate combined features so as to use existing features more effectively. We also propose a novel clustering algorithm which uses both linguistic knowledge and statistical knowledge. We build a coreference resolution system based on the proposed method and evaluate its performance from three aspects: the contributions of active learning; the effects of different clustering algorithms; and the resolution performance of different kinds of NPs. Experimental results show that additional performance gain can be obtained by using active learning method; clustering algorithm has a great effect on coreference resolution's performance and our clustering algorithm is very effective; and the key of coreference resolution is to improve the performance of the normal noun's resolution, especially the pronoun's resolution.

This paper investigates a machine learning approach for identification of temporal relation between events in Chinese text. We proposed a temporal relation annotation guideline [3] and constructed temporal information annotated corpora. However, our previous criteria did not deal with various uses of Chinese verbs. For supplementing the previous version of our criteria, we introduce attributes of verbs that describe event types. We illustrate the attributes by the different examples of verb usages. We perform an experiment to evaluate the effect of our event type attributes in the temporal relation identification. As far as we know, this is the first work of temporal relation identification between verbs in Chinese texts. The result shows that the use of the attributes of verbs can improve the automatic temporal relation extraction accuracy.

Bilingual dictionary construction is a time-consuming job; therefore many studies have recently focused on automatically constructing bilingual dictionaries from bilingual texts. In this paper, we propose two novel approaches called dynamic window and tokenizer based on statistical machine transliteration model to efficiently extract English-Chinese transliteration pairs from parallel corpora. Our approaches achieve a significant improvement over previous methods without using resources such as a Chinese segmentor, pronunciation dictionaries; hence these approaches can be readily applied to other language pairs.

We study an adaptive learning framework for phonetic similarity modeling (PSM) that supports the automatic acquisition of transliterations by exploiting minimum prior knowledge about machine transliteration to mine transliterations incrementally from the Web. We formulate an incremental learning strategy for the framework based on Bayesian theory for PSM adaptation. The idea of incremental learning is to benefit from the continuously developing history to update a static model towards the intended reality. In this way, the learning process refines the PSM incrementally while constructing a transliteration lexicon at the same time. We further demonstrate that the proposed learning framework is reliably effective in mining live transliterations from Web query results.