Recent advances in Large Language Models (LLMs) have achieved remarkable breakthroughs understanding and responding to user intents. However, their performance lag behind general use cases some expertise domains, such as Chinese medicine. Existing efforts incorporate medicine into LLMs rely on Supervised Fine-Tuning (SFT) with single-turn distilled dialogue data. These models lack the ability for doctor-like proactive inquiry multi-turn comprehension cannot align responses experts'...

Just-noticeable distortion (JND), which refers to the maximum that human visual system (HVS) cannot perceive, plays an important role in perceptual image and video processing. In comparison with JND estimation for images, of profile needs take into account temporal HVS properties addition spatial properties. this paper, we develop a spatio-temporal model estimating discrete cosine tranform domain. The proposed incorporates contrast sensitivity function, influence eye movements, luminance...

Background Individuals diagnosed with type 2 diabetes mellitus (T2DM) commonly exhibit elevated lipid levels and an increased body mass index (BMI). The impact of BMI on the efficacy statins in reducing among diabetic patients remains uncertain. We aim to evaluate whether will affect lipid-lowing effects T2DM. Methods In this retrospective analysis, we screened T2DM who were prescribed underwent a 1-year outpatient follow-up recorded electronic medical record system. Patients stratified into...

Machine transliteration is an important Natural Language Processing task. This paper proposes a Noisy Channel Model for Grapheme-based machine transliteration. Moses, phrase-based Statistical Translation tool, employed the implementation of system. Experiments are carried out on NEWS 2009 Transliteration Shared Task English-Chinese track. back studied as well.

Recent advances in Large Language Models (LLMs) have achieved remarkable breakthroughs understanding and responding to user intents. However, their performance lag behind general use cases some expertise domains, such as Chinese medicine. Existing efforts incorporate medicine into LLMs rely on Supervised Fine-Tuning (SFT) with single-turn distilled dialogue data. These models lack the ability for doctor-like proactive inquiry multi-turn comprehension cannot align responses experts'...

Text normalization is an important component in text-to-speech system and the difficulty text to disambiguate non-standard words (NSWs). This paper develops a taxonomy of NSWs on basis large scale Chinese corpus, proposes two-stage disambiguation strategy, finite state automata (FSA) for initial classification maximum entropy (ME) classifiers subclass disambiguation. Based above taxonomy, approach achieves F-score 98.53% open test, 5.23% higher than that FSA based approach. Experiments show...

Emotion classification of text is very important in applications like emotional text-to-speech (TTS) synthesis, human computer interaction, etc. Past studies on emotion focus the writer's state conveyed through text. This research addresses reader's emotions provoked by The documents into reader categories has novel applications. One them to integrate a Web search engine allow users retrieve that contain relevant contents and at same time produce proper emotions. Another for sites organize...

The DimABSA task requires fine-grained sentiment intensity prediction for restaurant reviews, including scores Valence and Arousal dimensions each Aspect Term. In this study, we propose a Coarse-to-Fine In-context Learning(CFICL) method based on the Baichuan2-7B model in SIGHAN 2024 workshop. Our improves accuracy through two-stage optimization process. first stage, use fixed in-context examples prompt templates to enhance model's recognition capability provide initial predictions test data....

Character is one of the three elements a novel, and conversation an important way to describe characters. The personality, emotion, interpersonal relationships characters are reflected in conversations. Thus, extracting conversations, speakers other from novels crucial for character analysis content understanding. We start with Jin Yong’s novels, annotate largest Chinese corpus speaker identification 9721 quotes, analyze language styles different based on quotes. then propose machine...

Aspect-based sentiment analysis (ABSA) aims to judge the polarity of specific aspects in text reviews, and is a fine-grained task. In current e-commerce era, ABSA based on user reviews great significance consumers, producers sellers. order make full use dependency information text, we propose graph convolutional network model for ABSA. Two networks are used encode edge tag respectively, then biaffine module realize interaction between two. The experimental results show that proposed...

It is still an open problem to efficiently explore BERT for contextual question answering, such as conversational reading comprehension (CRC). Previous work on using CRC does not deeply integrate conversation history into the architecture of BERT. In order make better use CRC, in this paper, we propose HisBERT (BERT with history) that consists two parallel units and divides three blocks history. Additionally adversarial training disturb word embedding layer, so increase robustness proposed...

Aspect-based sentiment classification (ASC) is a task to determine the polarities of specific aspects in review. Syntactic information like dependency relation has been proven effective when extracting features. On other hand, multiple semantic segments review may influence polarity. Thus, we propose neural network based on and structured attention (DRSAN) fuse both features with different mechanisms. To verify performance DRSAN, build Chinese Mobile Phone Review (CMPR) dataset. our...

In the past few years, it has become increasingly popular to analyze information obtained develop services by conducting a decentralized survey of private data for specific populations. Privacy security requirements providers force operators implement reasonable privacy protections. But increasing investment in protection will also lead decline operator revenue. this case, need ensure and users while ensuring sustainability customized services. To end, We study relationship between...

Character is one of the three elements a novel, and conversation an important way to describe characters. The personality, emotion, interpersonal relationships characters are reflected in conversations. Thus, extracting conversations, speakers other from novels crucial for character analysis content understanding. We start with Jin Yong's novels, annotate largest corpus speaker identification 9721 quotes, propose machine learning-based method, design feature templates showing good...

Selectional preference (SP) is an important semantic knowledge. It can be used in various natural language processing tasks, including metaphor computing, lexicon building, syntactic structure disambiguation, word sense role labeling, etc. However, handcrafted SP knowledge not meet the requirement of large scale real text processing. Based on noun taxonomy How Net, this paper proposes a statistical and knowledge-based method to automatically induce Chinese SP. Preliminary experimental...

Emotion Cause Pair Extraction in Conversations (ECPEC) is a challenging new task the field of sentiment analysis. Its objective to extract emotion utterances and their corresponding cause from conversation. Most recent studies have adopted end-to-end approaches handle this task. However, it difficult for these fully address issue label sparsity data. Thus, we introduce Machine Reading Comprehension (MRC) framework with Position-aware Graph Convolutional Network (GCN) leverage dialogue...

Multi-domain aspect-based sentiment analysis (ABSA) seeks to capture fine-grained across diverse domains. While existing research narrowly focuses on single-domain applications constrained by methodological limitations and data scarcity, the reality is that naturally traverses multiple Although large language models (LLMs) offer a promising solution for ABSA, it difficult integrate effectively with established techniques, including graph-based linguistics, because modifying their internal...