id
stringlengths 7
12
| sentence1
stringlengths 6
1.27k
| sentence2
stringlengths 6
926
| label
stringclasses 4
values |
|---|---|---|---|
train_99500 | When the requester uses factor F5, he/she assumes that explaining the reasons to the requestee and requestee's potential understanding of reasons of request may increase the likelihood of the fulfilment of a request. | our aim was to make an analysis of request texts written in English, Spanish and Slovak language, where we examined the occurrence of keywords, in our case the factors of politeness in mother tongue (Slovak) and in foreign languages (English and Spanish). | neutral |
train_99501 | The algorithm employs a human judge to prevent semantic drift in the iterative process of a bootstrapping algorithm. | for example, given a few instances of car vehicles "Prius", "Lexus" and "Insight", the task outputs new car instances such as "Corolla", "Civic", and "fit". | neutral |
train_99502 | In the previous studies on bootstrapping, if the method obtains "new type of X" as the pattern of car category, then the method does not have a mechanism to reject incorrect instances such as "iPhone". | for example, we can obtain ontological information between semantic categories easily from existing resources such as Wikipedia. | neutral |
train_99503 | This fact indicates that sibling relation and exclusivity between categories improve the set expansion accuracy. | we demonstrate the effectiveness of sibling relations in set expansion on the dataset in which instances and sibling relations are extracted from wikipedia in a semi-automatic manner. | neutral |
train_99504 | After the method secures new instances of each category, the method proceeds to the next iteration. | investigating the impact of choosing sibling groups would be an interesting future direction of this research. | neutral |
train_99505 | Recent advances in NLP tools argue that linguistic features are useful for readability classification. | for example: the word biodegradable will be harder to pronounce, spell and understand for children at a lower grade level. | neutral |
train_99506 | Rarely occurring relations (<5 times) have not been considered in the table. | a wine bottle is a type of bottle, glass furniture is a type of fur- Relation annotator 1 annotator 2 Part-Whole 184 90 Type 178 192 Purpose 132 122 Source 70 78 Property 25 22 Hypernymy 25 30 Location 34 30 Topic 12 8 Theme 8 8 Temporal 5 5 Table 4: Distribution of annotated data among the relations. | neutral |
train_99507 | They calculated the similarity between a given test instance and every training instance in the training set, and predicted the semantic relation of the most similar training instance. | since the action features are verbs and cannot act as modifiers in a nominal compound, we extract the descriptive features D of the node corresponding to every A A and append D and D to the list of potential modifiers M . | neutral |
train_99508 | As standard machine learning pipeline, we perform EM on training set, tune parameters on development set, and report the result of selected model on test set. | the second constituent/distituent factors (F c:x and F d:x ), are the concatenation of context features and constant feature. | neutral |
train_99509 | Since the word vectors have non-negative components, the range is between 0 and 1, where a value of 0 indicates that the vectors are orthogonal or dissimilary and a value of 1 indicates that the vectors are parallel or similar. | in this work, we make a choice of defining the cooccurence relation as the joint cooccurence of the word within a short window of text, by the principle of occam's razor. | neutral |
train_99510 | The strategy used in evaluating the phrasal similarity score from the component token similarity scores is critical in deciding the overall performance of the MEANT metric, as role fillers are often phrases. | the robustness experiments were conducted across various data sets following NISt Met-ricsMatr protocols, showing higher Kendall correlation with human adequacy judgments against BLEU, MEtEOR (with and without synsets), WER, PER, tER and CDER. | neutral |
train_99511 | Previous research on the classification of these nouns is based on their semantic categories (Han 2004(Han , 2010bLiu 2004;Wang 2010;Zhong 2010). | this is the similar to entity nouns. | neutral |
train_99512 | It is because hou coerces the NP into obtaining an event denotation, one which is available from the NP's qualia structure. | (3)Type Coercion: the type a function requires is imposed on the argument type. | neutral |
train_99513 | (2011) divide them into "numeral classifiers" and "measure specifiers" depending on their functions, which will be shown in 2.1. | bond and Paik, 1997;Huang and Ahrens, 2003;Kageyama et al., 2011;Mano and Yonezawa, to appear), and there seems to be still room for argument. | neutral |
train_99514 | This paper is concerned with characterizing psych-predicates in Korean and possibly in Japanese in the GL spirit. | inanimate subjects in a transitive causative sentence are not fully acceptable in Korean. | neutral |
train_99515 | the ox-Top father-Nom sell-Past-Dec 'The ox, father sold it.' | father-Nom sell-Adn KES-Top ox-be-Dec 'What father sold is an ox.' | neutral |
train_99516 | They can be represented by some verbs of arousal, being emitted (by), or result (or leaving behind), etc. | (1) cause-effect relation with sensory head noun [sayngsen-i tha-nun] naymsay fish-Nom burn-Adn smell 'the smell that comes from fish burning' (2) cause-effect relation with non-sensory head noun [thayphwung-i cinaka-n] huncek typhoon-Nom pass-Adn trace 'the trace left after a typhoon hit' (3) cause-effect relation with non-natural phenomenon [apeci-ka so-lul phal-un] ton father-Nom ox-Acc sell-Adn money 'the money obtained by selling an ox' It is observed that there exists a semantic cause-effect relation holding between the GRC and its modifying head noun: the content of the adnominal GRC constitutes cause and the denotation of its head noun effect. | neutral |
train_99517 | But this analysis does not seem to have any repertoire of deep explanatory devices for the above state of affairs. | the paper shows that this incomplete realization of the cause-effect relation can be fully recovered from the lexical-semantic(-pragmatic) information specified under the GL framework. | neutral |
train_99518 | KES-Top money-be-Dec 'What father earned by selling an ox is money.' | argument coherence of identity between the agent 'father' of the ox-selling causal event that appears in the adjunct clause and the agent 'father' of the money-earning effect that appears in the event phrase or clause is well observed (Pustejovsky 1995). | neutral |
train_99519 | It is just a coincidence that the SI goes into the relative clause. | c. John won the lottery or made some easy money. | neutral |
train_99520 | SIs are inferences based on the maxim of quantity by Grice (1975). | the SI that it is not the case that Mary believes John broke many/all glasses does not mean that Mary believes John did not break many/all glasses. | neutral |
train_99521 | A DRS is supported by an information state iff each pair of a possible world and an assignment supports each condition in the DRS. | this is a deviation from the standard DRS, in which each DRS is interpreted with respect to a model. | neutral |
train_99522 | I do not have to follow the Neo-Gricean tradition, because it is assumed that a SI arises by comparing the meanings of alternative statements. | watching TV or playing games is understood as the lower limit for failing math. | neutral |
train_99523 | Thus the global SI has the effect of a local SI. | it is the first maxim of quantity that is relevant to Sis. | neutral |
train_99524 | Variables like Z and k are introduced by presuppositions and are supposed to be bound by Z and k respectively. | this effect is obtained in various ways. | neutral |
train_99525 | Similarly, in (9), if either of the two requirements is sufficient to get a grade, it is more plausible to assume that no students satisfied both requirements. | contextual information should be put aside when we discuss the way that SIs are calculated. | neutral |
train_99526 | Traditionally it has been claimed that C/PM do not co-occur on N in Mandarin. | in [1×1], either the multiplier (Num) or the multiplicand (C/PM) can be omitted without changing the result. | neutral |
train_99527 | 5 The selection of a term is pragmatically determined. | while the entry of no in Cann et al. | neutral |
train_99528 | Backward-direction Markov processing starts with a noun, and extends to the left direction based on a bi-directional Markov dictionary as shown in Figure 3. | the score varies inversely with the number of morphemes in the snippet fragment, and even gets N i times lower when more than 2 potential nominatives might exist. | neutral |
train_99529 | The relationships for oh and dui in mcdc05 are consistent with the results for mcdc01, with female oh and male dui having the higher relative frequencies, as presented earlier. | oh, dui, and umhum each has its unique different functions and occurs under different informational environments. | neutral |
train_99530 | The Taiwan Mandarin dataset contains only very long speaker turns, thus reducing the number of shorter prosodic units which are more often produced in interactive conversational speech. | 1 We had to decide which linguistic information and which part from the full corpora to include in our joint dataset. | neutral |
train_99531 | Other measures need to be introduced (Pevzner and Hearst, 2002;Fournier and Inkpen, 2012) if one wants to measure a different aspect of the segmentation agreement. | it is striking to see in figure 3 that more than half of the time and for both language discourse units are providing the starting and ending boundaries for the prosodic units. | neutral |
train_99532 | (Cortes, 2004;Hyland, 2008a;Lorés, 2004;Martín, 2002;Swales, 1990;Wang & Chan, 2011;Wang & Kao, 2012). | taking the contribution of the previous studies, this current reserach sets out to explore the variation of tense within the reporting verbs among the transitions of moves via the structural analysis in JA abstracts. | neutral |
train_99533 | Finally, we conclude this paper and discuss future work. | section 2 presents the main framework of our ChNWI approach. | neutral |
train_99534 | From this consideration, for a Chinese string S, if its left adjacent category value or its right adjacent categories are larger than a predetermined threshold, which means that the string S is loose with its context and it is higher possibility of being a Chinese new word. | this work is sponsored by the grant from the Shanghai Science and technology Foundation (No. | neutral |
train_99535 | For the tagging bias feature, the scoring function will be: is selected according to which tagging bias probability factor is used and whether 1 i i c is seen. | furthermore, prefix/suffix related features were claimed to be useful for CWS in some previous works (Tseng et al., 2005;Zhang et al., 2006). | neutral |
train_99536 | It is observed that the unseen ratio for qs is much larger than that for qf. | having considered that the surrounding context might be unseen for the testing instances, we introduce four different kinds of tagging bias probabilities as follows (and they will be trained in parallel for each character in the training-set): . | neutral |
train_99537 | Although the character and character co-occurrence features (adopted in most current approaches) are able to partially characterize the internal structure of words (Sun, 2010), and some OOV words are indeed correctly handled when compared to pure wordbased approaches (Zhang et al., 2003;Gao et al., 2005), suffix related errors still remain as an important type of errors. | in comparison with the discriminative model, the generative model has the drawback that it cannot utilize trailing context in selecting the position tag (i.e. | neutral |
train_99538 | Nonetheless, we cannot distinguish suffixes from those non-suffixes by just checking each character because whether a character is a suffix highly depends on the context. | in their works, the prefix/suffix features are just a part of adopted features. | neutral |
train_99539 | Rather, human beings perceive speech by generalizing others' gestures, no matter he or she have such knowledge of gesture. | if the latter is true, the learnability of L2 speakers in one sound may be discovered to be different in different gestural environments. | neutral |
train_99540 | Especially, experienced learners who are considered near-native in proficiency will often establish stable intermediate categories in an audio-based learning model, the most widely renowned being the Speech Learning Model (SLM, Flege, 1987;Flege et al., 2003). | all participants were right-handed with no reported hearing or motorcontrol defects. | neutral |
train_99541 | This shows that acoustic sound is not a basis for L2 speech perception and the results supports the motor theory that speech is not perceived through sounds exclusively. | this model would bring fine-grained gestural percepts and frequency-based normalizing process of category formation together. | neutral |
train_99542 | The participants were asked to circle the correct word on the answer sheet. | for alveolar clusters (tr-and dr-), previous studies showed that considerable affrication was a feature of their production (Lan and Oh, 2012). | neutral |
train_99543 | The constitution of Chinese tones is not determined only by pitch level, but also by transition patterns. | in this study, the errors made by the 10 subjects were mostly found in tonal patterns 1-3, 2-1, 2-3, 3-2 and 4-3. | neutral |
train_99544 | When tone 3 is in the initial position, it is likely to be heard as tone 2 or tone 4. | in the same way, if the first is with low pitch, the second must be with high pitch. | neutral |
train_99545 | comfortable to sofa) and functions (e.g. | paradigmatic relations may be more salient for nominal concepts, whereas syntagmatic relations, especially nounverb and noun-adjective associations, are more significant for the learning and memory of verbs and adjectives. | neutral |
train_99546 | For the experiment by the human judgement, we concluded that the types of "Word choice (SEM)", "Missing (OM)" and "Unnecessary (AD)" can be included in any other error types, which causes the confusion regardless of the machine or the human classification. | we will take those aspects into the future trial. | neutral |
train_99547 | To extend the line of frog-story-based research and to replicate previous findings based on causal networks, the present study explored how Mandarin-speaking children maintain narrative coherence by posing the following research questions. | the reverse pattern, however, is shown for the density of causal-chained events. | neutral |
train_99548 | Similarly, Stein and Albro (1997) suggested that causal reasoning is required to organize content and structure coherently. | in the analysis based on 'plot' components, researchers found developmental increases in children's ability to establish global plotline (Aksu- Koç and Tekdemir, 2004;Berman and Slobin, 1994). | neutral |
train_99549 | Such research has been conducted across the eleven sub-corpora by Gilquin and Granger (2011;forthcoming). | being one of the sub-corpora of LINDSEI, without doubt, increases the visibility of Taiwan in international academia and contributes to the research on spoken English. | neutral |
train_99550 | Each of them is grounded on a particular framework with certain assumptions, one more suited for certain applications than the others. | annotated on selected texts from the Penn Treebank, the corpus was built by a collaborative team on web-based annotation platform with several advanced features. | neutral |
train_99551 | A number of linguistic resources annotated with shallow semantic roles have been produced over the years. | in the sentences: John broke the window, and The window broke, although the window is the syntactic object in the first and subject in the second, it is given the same argument label. | neutral |
train_99552 | which have to be used for further specification of the relative position of the figure. | tai (tai 1989(tai , 1992(tai , 1993, seems to be a natural continuation of the descriptive tradition. | neutral |
train_99553 | Figure 1: Structure of wū-lǐ ('in the room') The resulting structure can be used in two contexts: on the one hand, they can be used as prepositional complements in locative adjuncts, as illustrated in (3b). | • Valence Principle: in a headed phrase, for each valence feature F, the F value of the mother is determined as follows: -If the valence list of F consists of synsem objects, its value corresponds to the head daughter's F value minus the SYNSEM values of its sisters. | neutral |
train_99554 | A Trale grammar can be distributed between an arbitrary number of files, different files containing sets or subsets of linguistic generalizations of a certain type. | in the following exposition, we often use simplified structures for purposes of illustration in order to ease understanding by non-Chinese speakers. | neutral |
train_99555 | Second, Chinese characters that are still popularly used in modern Chinese may be only phonetically borrowed by Chữ Nôm. | for example, the Vietnamese language in Vietnamese grammar order should not be Vietnamese language (Việt Nam tiếng) but language Vietnamese (tiếng Việt Nam ). | neutral |
train_99556 | We choose one Buddhist scripture as our data set for evaluation from the Chinese Buddhist Canon maintained by Chinese Buddhist Electronic Text Association (CBETA). | there are two types of parsing errors to a token in a dependency parse tree. | neutral |
train_99557 | The probability of a sequence of m words can be transformed into a conditional probability: In practice, we can assume the probability of a word only depends on its previous word (bi-gram assumption). | our approach comprises the conditional random fields method with designed features which are suitable to identify transliteration characters. | neutral |
train_99558 | Due to the time limitation, we only focused on analyzing parsing errors that cause reordering issues on Vb-H candidates while defining the error patterns. | recall and precision are widely used measurements in many research fields, sucn as information retrieval and information extraction. | neutral |
train_99559 | Indeed, DPC heavily relies on detecting verbal blocks that are candidates for reordering, and systems that use the same strategy should choose POS taggers that display high accuracy of VV recognition. | the language pair under consideration is Chinese-to-Japanese, which despite their common roots, it is a well known language pair for their different sentence structure. | neutral |
train_99560 | Table 2 shows the statistics of all selected sentences in five genres respectively. | in the example of Figure 3, dependenterror counts are for VV, 2 errors, and PN, NN, CD, PU each 1 error. | neutral |
train_99561 | As introduced in Section 2.1, DPC first identifies Vb, RM-D, and then reorders necessary words. | for scenario 1, the benchmark is manually reordered Chinese sentence on the basis of Japanese reference. | neutral |
train_99562 | A proper subset of 'every child' cannot be also described as 'every child', for the reason that if a set y is the proper subset of a larger set x, y is necessarily smaller and thus does not include at least one of the members in x. | since these subevents are homogeneous, the whole reading event is considered atelic. | neutral |
train_99563 | Section 4 briefly discusses ditransitive causatives and their related forms. | since the pattern observed in these examples cannot be predicted, as shown in Table 1, it is clear that the causativization approach should be excluded even in cases where it might appear to be appropriate. | neutral |
train_99564 | Given, moreover, that it is implausible to derive the transitive from the ditransitive causative considering the morphological makeup, the common base approach is the only approach that is left to account for the contrast in (42). | (27) a. Taroo-ga sira-o kir-Ø-ta (>kit-ta), T.-NOM white-ACC √detach-CAUS-PST Ziroo-mo {soo si-ta/ kit-ta} Z.-also so do-PST/ (see above) 'Taroo dissembled. | neutral |
train_99565 | In an attempt to answer this question, this paper examines the three approaches to the causative alternation by utilizing the possibility of having idiomatic interpretations as a probe into the syntactic structure. | it predicts that, if an idiomatic interpretation is possible with a causative, then it should be possible with its inchoative counterpart, as shown in Table 2 below. | neutral |
train_99566 | More details concerning the corpus can be found in Cheng et al. | the values of v(t) curves are from the original un-normalised F 0 contours (and are thus in units of semitones per second) before themselves being normalised in time. | neutral |
train_99567 | Therefore the result of relative pitch height could be noted as realized in one of the following situations: R1 is the higher segment of the DR (Higher R1) R2 is the higher segment of the DR (Higher R2) R1 and R2 are perceived as realized in the same pitch height (Same) No comparable result could be yield as for which segment of R1/R2 is higher (No result) From Figure 1 it is demonstrated that for most of the DR instances in both languages R2 is perceived as relatively higher when executing the repair (For French recyclings: 64/150, 42.6%; Mandarin: 79/150, 52.7%). | part of the result did suggest that, other than direct repetition (i.e. | neutral |
train_99568 | In a similar vein, Kreyer (2003) investigates corpus data which also allow for a possible alternation between genitive and ofconstruction (e.g., the chairman of the committee and the committee's chairman) and shows that processability and degree of human involvement are two crucial factors influencing speakers' selection of the constructions. | category include a number of fixed expressions (e.g., point of view) and irrelevant structures (e.g., 'demonstrated approval of them' EF3660) which were excluded from further analysis (see Appendix I for raw scores). | neutral |
train_99569 | Among the four groups, only the N2 heads with N1 'measure' nouns reach statistical significance for demonstrate. | we would expect an evaluation embedded in the of-nominals by means of modified N1 or factual nouns found in the 'support' group. | neutral |
train_99570 | Such association measures indicated how much more often than chance the NP co-occurred with the spatial particle. | the present study would like to investigate whether bipolar spatial particles (e.g., up/down) on the same spatial dimension (e.g., vertical axis) exhibits a symmetrical extension to similar sets of target domains in the real language use. | neutral |
train_99571 | A cluster of CFE markers, as listed above, all contribute to an overall optimum situation for expressing hypothetical events, and the more CFE markers that are present in the sentence, the higher degree of hypotheticality implied, with counterfactuality being obtainable as an inference or implicature deriving from the highest level of hypotheticality (Comrie, 1986). | if Manila weren't so far away, i'd send him to study there. | neutral |
train_99572 | At the first stage in that the markers are restrained to past time reference (a),imply couterfactuality in the strict sense (dependence on a condition known to be false)(b), are optional(c). | we cannot easily take all these strategies as markers for CFs in Chinese. | neutral |
train_99573 | In this case, the former interpretations were tentatively adopted by the sys-tem. | "(g) the framework for the event sequence model should determine how to achieve mutual reference to semantic information between different event sequences." | neutral |
train_99574 | Thus, there should be overlapped feature between sarcasm and irony based on the speaker awareness. | (3) Is the tweeter aware of his/her sarcastic or ironic tweets? | neutral |
train_99575 | Different from this, the tweets of second sense of irony is tagged to classify an event to be ironic. | (1-f) This play is amazingly good. | neutral |
train_99576 | This shows that ironic tweets can contain the tweets that are with speaker's awareness. | in computational linguistics irony and sarcasm are lack of quantitative distinction, but are viewed as the same in data processing (Elena 2012), which may with the danger that the true intention of the speakers on their opinions may be wrongly captured because psychological studies (Lee and Katz 1998) and qualitative study (Clift 1999) have discovered that sarcasm is the real aggressive one, but irony is not necessarily meant to attack. | neutral |
train_99577 | #sarcasm (1-i) you always find a way to ruin my night! | in ironic tweets though the attack on specific target is rarely few, it does exist. | neutral |
train_99578 | Different from the previous studies that are aware of the user and product of the review, our model does not assume the training data to contain the reviews written by the same user of test reviews or written on the same product of test reviews. | since the two users are lenient and the first product is relatively popular (these characteristics are captured by our proposed method), these two reviews should still be given positive labels. | neutral |
train_99579 | We suppose that the characteristics of news articles are the reason why nested-source and direct-speech (type) are the main features in the experimental results. | the following Figure 3 shows the best results and the experimental result of using all attributes. | neutral |
train_99580 | This kind of methods could obtain high precision, while suffer from low recall. | automatically creating crosslingual links between Chinese and English Wikipedia would be very useful in information flow and knowledge sharing. | neutral |
train_99581 | Wikipedia knowledge base is a collection of collaboratively written articles, each of which defines a specific concept. | we extract all n-grams of size 1 to 5, because n-grams subsume most key term candidates, which could obtain a high recall. | neutral |
train_99582 | All the transliteration systems return transliterations, good enough to understand the text, which is strengthened from the evaluators' score as well as from the character-level ac-curacies. | in this work, we have modelled the task of transliteration as a noisy channel model with minimum error rate training (Och, 2003). | neutral |
train_99583 | All these sentences were transliterated by all the 7 transliteration systems and the results of each were given to several evaluators 8 to rate the sentences on the scale of 0 to 4. | such a statistical modelling needs an ample amount of data for training and testing. | neutral |
train_99584 | To measure the similarities between the transliterated word w x and target candidate word w y , the Dice coefficient (Dice, 1945) is used. | we also extracted sentences that contain out-of-vocabulary words (813 lines) from the test set. | neutral |
train_99585 | As described in , the correspondence-based transliteration model (Oh and Choi, 2002) is also considered as a hybrid approach. | there are 50 (10.00%) and 55 (11.00%) entries contain three or more character errors. | neutral |
train_99586 | where the insert_tree(.) | the order of frequent items in FPtree storing the compressed frequent pattern set is the same as that of the original FP-tree (i.e. | neutral |
train_99587 | Primary tags are tags intended to be disambiguated and evaluated, and differences in e.g. | in order to isolate the relative contributions of promoting, demoting and rule killing, these 6 Further continuous 0.05 step variation was performed, but followed the general tendency and were left out in table 2. | neutral |
train_99588 | Secondly, the order of rules is also important. | adding new top/bottom-sections produced the highest recall gains (0.441 for DKRsS), but these did not translate into corresponding Fscore gains. | neutral |
train_99589 | With an F-score around 96% this corresponds roughly to a 7.5 % lower error rate in relative terms. | the question is therefore, whether a handcorrected gold corpus and machine-learning techniques could be used to improve performance by data-driven rule ordering or rule adaptation, applied to existing, manual grammars. | neutral |
train_99590 | Next, we apply random forests classification experiments using newspapers, editorial dates, and ideology types as classes in order to examine the classification performance and important features of these experiments. | this study analyzed the differences among newspaper editorials, focusing on five newspapers, two editorial dates, and ideology types. | neutral |
train_99591 | 12CP When susi-o 'sushi-ACC' is encountered, it is identified as an NP that has no theta-role assigned. | otherwise, word order would be rearranged. | neutral |
train_99592 | It has been generally assumed that a violation of island constraints indicates that the relevant syntactic phenomena involves movement. | (15) would have the reading that Taro knows the rumor that someone respects Mr. Tanaka, which is different from what is expected. | neutral |
train_99593 | Tables 2 and 3 show the search results and user satisfaction, respectively. | twenty-eight taiwanese learners of Spanish who studied in the Department of Foreign Languages and Literature participated in the survey. | neutral |
train_99594 | A contour like LH in Japanese might have disappeared so that Tone 3 in Chinese became the least favored choice for Japanese. | japanese pitch-accent and Korean tones? | neutral |
train_99595 | For each verb in the list, we stop traversing if either we exhaust the nodes dominated by the verb or find another verb in its dominance. | although, in STEP 1 we will accurately include the complement clause as a part of the main clause, we don't mark the scope of complement clause itself, if it is complex i.e., made of more than one clause. | neutral |
train_99596 | Among the two wordinternal elements, derivational affixes are more closely related to their roots than inflectional affixes to their stems. | bengkyou yoku/nagaku/... suru well/long time c) [bengkyou-to undou]-bakari/wa/... suru -and exercise Although it is not very natural for such independent words as yoku and nagaku to come between the two elements, as we can see in (b), delimiters like -bakari and -wa are allowed as in (a). | neutral |
train_99597 | From a computational point of view, the sheer number of ha-expressions in Korean makes the popular misanalysis more difficult to maintain. | in this section, we will examine the behavior of the verbal element -(nu)n-. | neutral |
train_99598 | When we deal with computational systems, we have to consider the understanding process and the productions process separately, just as the two areas of speech recognition and speech synthesis show. | 1 as we can see shortly, ha-itself is an independent word and [kongpwu ha-] is a phrase. | neutral |
train_99599 | In the remaining sections of this paper, we will focus on only one example from each of these two types of misanalyses: the "light verb" ha-and the assumed present tense marker -(nu)n-. | their existence has 2 The abbreviations used for grammatical terms in this paper are as follows. | neutral |
Subsets and Splits
No saved queries yet
Save your SQL queries to embed, download, and access them later. Queries will appear here once saved.