Cross-domain constituency parsing is still an unsolved challenge in computational linguistics since the available multi-domain constituency treebank is limited.We investigate automatic treebank generation by large language models (LLMs) in this paper.The performance of LLMs on constituency parsing is poor, therefore we propose a novel treebank generation method, LLM back generation, which is similar to the reverse process of constituency parsing.LLM back generation takes the incomplete cross-domain constituency tree with only domain keyword leaf nodes as input and fills the missing words to generate the cross-domain constituency treebank.Besides, we also introduce a span-level contrastive learning pretraining strategy to make full use of the LLM back generation treebank for cross-domain constituency parsing.We verify the effectiveness of our LLM back generation treebank coupled with contrastive learning pre-training on five target domains of MCTB.Experimental results show that our approach achieves state-of-theart performance on average results compared with various baselines.

Knowledge transfer is investigated in various natural language processing tasks except cross-domain constituency parsing. In this paper, we leverage heterogeneous data to transfer cross-domain and cross-task knowledge to constituency parsing. Concretely, we first select language modeling, named entity recognition, CCG supertagging and dependency parsing as auxiliary tasks and collect the corpora of these tasks covering various domains as cross-domain and cross-task heterogeneous data. Second, we exploit three types of prefixes: shared, task and domain prefix, to merge cross-domain and cross-task data and decompose the general, task and domain representation in the pretrained language model. Third, we convert the data formats of multi-source heterogeneous datasets and loss objectives of the auxiliary tasks into a consistent formalization closer to constituency parsing. Finally, we jointly train the model to transfer task and domain knowledge to cross-domain constituency parsing. We verify the effectiveness of our proposed model on five target domains of MCTB. Experimental results show that our knowledge transfer model outperforms various baseline models, including conventional chart-based and transition-based parsers and the current large-scale language model for zero-shot and few-shot settings.

Self-training has proven to be an effective approach for cross-domain tasks, and in this study, we explore its application to cross-domain constituency parsing.Traditional self-training methods rely on limited and potentially lowquality raw corpora.To overcome this limitation, we propose enhancing self-training with the large language model (LLM) to generate domain-specific raw corpora iteratively.For the constituency parsing, we introduce grammar rules that guide the LLM in generating raw corpora and establish criteria for selecting pseudo instances.Our experimental results demonstrate that self-training for constituency parsing, equipped with an LLM, outperforms traditional methods regardless of the LLM's performance.Moreover, the combination of grammar rules and confidence criteria for pseudo-data selection yields the highest performance in the crossdomain constituency parsing 1 .

Predicting tumor malignancies is an important but difficult task. For many tumors, especially neural and endocrine tumors, traditional pathological and histological analyses often can not effectively distinguish benign from malignant tumors. Developing synergistic bioinformatics and computational intelligence system is effective, because deterministic cancer markers do not always exist in individual patients. We proposed a parallel paradigm of cancer and use a number of ensemble methods including boosting, bagging and consensus networking, and have designed a novel classification scheme that advantageously combines several computational intelligence algorithms such as the variants of self-organizing feature map (SOFM) algorithms and the maximum contrast tree (RMCT) algorithms. Boosting and bagging have been advantageously combined. When all of the above are integrated into one synergistic intelligent medical decision system, the prediction power for the task has been significantly boosted. The system and features are validated by diagnosing new patients and by a number of laboratory molecular biology measurements. The outcomes of the research have improved cancer diagnosis and treatment planning, and may lead to diagnose microscopic diseases and better understanding of human genome mechanisms relating to malignant transformation

BACKGROUND: Many proteomics initiatives require integration of all information with uniformcriteria from collection of samples and data display to publication of experimental results. The integration and exchanging of these data of different formats and structure imposes a great challenge to us. The XML technology presents a promise in handling this task due to its simplicity and flexibility. Nasopharyngeal carcinoma (NPC) is one of the most common cancers in southern China and Southeast Asia, which has marked geographic and racial differences in incidence. Although there are some cancer proteome databases now, there is still no NPC proteome database. RESULTS: The raw NPC proteome experiment data were captured into one XML document with Human Proteome Markup Language (HUP-ML) editor and imported into native XML database Xindice. The 2D/MS repository of NPC proteome was constructed with Apache, PHP and Xindice to provide access to the database via Internet. On our website, two methods, keyword query and click query, were provided at the same time to access the entries of the NPC proteome database. CONCLUSION: Our 2D/MS repository can be used to share the raw NPC proteomics data that are generated from gel-based proteomics experiments. The database, as well as the PHP source codes for constructing users' own proteome repository, can be accessed at http://www.xyproteomics.org/.