Corpus. The Trustees may, in their discretion, pay or apply to or for the benefit of Primary Beneficiary such additional amounts from corpus, in such manner and at such intervals and in such amounts as the Trustees in their discretion from time to time shall deem requisite or desirable; provided, however, in the case of each payment of corpus, the Trustees shall first determine that such payment or application of corpus is reasonably necessary to promote Primary Beneficiary's education, support, maintenance, or health; or is reasonably necessary for the reasonable support and comfort of Primary Beneficiary; or is reasonably necessary to enable Primary Beneficiary to maintain her accustomed standard of living; or is reasonably necessary to meet an emergency. In making such determination under this subsection 2b, the Trustees shall take into consideration any other income (other than capital gains), or property known to the Trustees which Primary Beneficiary may have or enjoy from sources other than this trust estate. Section 3.
Corpus. The Trustees may, in their discretion, pay or apply to or for the benefit of any of Primary Beneficiary's blood issue, from time to time living, such additional amounts from corpus, in such manner and at such intervals and in such amounts, not necessarily in equal shares or amounts, as the Trustees in their discretion from time to time shall deem requisite or desirable. Provided, however, in the case of each payment or application of corpus, the Trustees shall first determine that such payment or application of corpus is reasonably necessary to promote the education, support, maintenance or health of such beneficiary; or is reasonably necessary for the reasonable support and comfort of such beneficiary; or is reasonably necessary to enable such beneficiary to maintain his accustomed
Corpus. Despite of all great work in QA, only two datasets are publicly available for machine comprehension that take dialogues as evidence documents. One is DREAM comprising dialogues for language exams with multiple-choice questions [23]. The other is FRIENDSQA containing transcripts from the TV show Friends with annotation for span-based question answering [28]. Since DREAM is for a reading comprehension task that does not need to find the answer contents from the evidence documents, it is not suitable for our approach; thus, FRIENDSQA is chosen. Each scene is treated as an independent dialogue in FRIENDSQA. Yang and Xxxx [28] randomly split the corpus to generate training, development, and evalua- tion sets such that scenes from the same episode can be distributed across those three sets, causing inflated accuracy scores. Thus, we re-split them by episodes to prevent such inflation. For fine-tuning (section 3.3), episodes from the first four seasons are used as described in Table 4.1. For pre-training (section 3.2), all transcripts from Seasons 5-10 are used as an additional training set. Set D Q A E Training 973 9,791 16,352 1 - 20 Development 113 1,189 2,065 21 - 22 Evaluation 136 1,172 1,920 23 - * Table 4.1: New data split for FriendsQA. D/Q/A: # of dia- logues/questions/answers, E: episode IDs.
Corpus. The Basic Transcription System for Japanese (BTSJ) contains 294 transcriptions of spontaneous Japanese speech, of which I have selected 36 for analysis (approximately 1/8 of the corpus). The files were divided into seven categories based on degree of familiarity. I followed the order of the corpus author’s categorization and tried to select a representative sample of transcripts from each of these main categories. For example, from the Teacher-Student category, there was at least one file of each type of conversation amongst genders, and so I selected one of each. On the other hand, from the Intimate category, there were only transcripts of female-female conversations, and so I selected three of those files. I also did not look at files that contained nonnative speakers. Each transcription file consists of two speakers with varying degrees of familiarity and differences in gender, as presented in Table 2. Degree of Familiarity # of Files # of Male-Male Conversations # of Female-Female Conversations # of Male-Female Conversations Intimate 3 0 3 0 Teacher-Student 3 1 1 1 Unacquainted 3 0 3 0 Table 2. Demographics of Transcription Files
Corpus. There were 2437 instances of ne occurring out of the 36 transcripts. Table 3 presents the distribution of the 2437 instances of ne. Placement of Ne Description # of Instances % Final Ne comes at the end of an utterance 1480 60.73 Post-Clause Ne comes after a clause (relative clauses, etc.), but not at the end of the utterance 444 18.22 Post-Interjection Ne comes after an interjection [eto (um), ano (uh), etc.] 164 6.73 Post- Topic/Subject Ne comes after the topic/subject 87 3.57 Post-Transition Ne comes after a transition word [soshitara (If so…), soreni (Moreover), etc.] 71 2.91 Alone Ne is the only word in the utterance 64 2.63 Other Includes ne coming after the direct object, postpositional particle/phrase, etc. 127 5.21
Corpus. The corpus data consisted of 2 hours of YM speech, containing 261 words repeated 6 times by 5 different native speakers. While a few words were produced with tonal changes that indicated negation or aspect marking, no overt segmental prefixes appeared on any word. A total of 7830 word repetitions were examined, comprising 27166 speech segments. All words were produced in citation form. Hand- labeling was done by the first author and compared to the forced alignments done by the P2FA and hmAlign models. Another set of labels provided both a phonemic transcription and encoded the location of the consonant or vowel in the word, e.g., C1 for the first consonant, V2 for the second vowel. This labeling allowed for the analysis of error reduction assessment by position in the word.
Corpus. VLSP 2018 - Named Entity Corpus ⃞ VLSP 2018 - Sentiment Corpus ⃞ Note: Choose one or multiple corpora by checking the corresponding box.
Corpus. The Settlor does herewith transfer and deliver unto the Trustee the sum of ($ ) DOLLARS as the initial corpus of this trust in accordance with Louisiana Revised Statutes, Title 8, Section 454A. The Settlor shall, as interment and/or entombment spaces in the Cemetery are sold or transferred, deliver to the Trustee ten per cent (10%) of the gross sales price received for each such space, and all funds so delivered shall be added to and form part of the corpus of this Trust.
Corpus. Trustee may pay to Settlor from the corpus of the trust from time to time such further amounts (even to the exhaustion of the trust) as in its discretion it deems necessary or advisable to properly maintain Settlor in the style to which Settlor is presently accustomed, and shall do so if Settlor becomes disabled. These payments may include amounts to or for the benefit of persons dependent on Settlor for support and premiums for life insurance on Settlor or those persons, whether or not the policies are payable to this trust.
Corpus. Trustee may invade the corpus of a separate trust for the benefit of the beneficiary if, in the Trustee's discretion, it is appropriate in order to:
