The realm of natural language processing has always been fascinated by the intricacies of sentence semantics. The emergence of context-aware word representations, especially from pre-trained models like ELMO and BERT, has revolutionized several semantic tasks including but not limited to, question answering, text categorization, and sentiment analysis. Despite these advancements, the integration of supplementary knowledge, particularly syntactic, to augment semantic comprehension in models remains an area ripe for exploration. This paper introduces the Syntactic Enhancement Network (SEN), a pioneering method for synergizing syntactic elements with established pre-trained language models. Our approach encompasses a dual-phase evaluation: initially, we delve into the syntactic augmentation of both RNN-based and Transformer-based language models; subsequently, we test our SEN's proficiency in two specific domains: the task of sentence completion and the extraction of biological relationships. The results are striking, with SEN achieving a stellar 91.2% accuracy in sentence completion—surpassing baseline models by a substantial 37.8%. In the context of biological relation extraction, SEN demonstrates competitive prowess with a 75.1% F1 score.

In this paper, we propose a novel approach named EmbedHDP to enhance the evaluation models used to assess sentence suggestions within nursing care record applications. The key focus is determining whether these suggestions garner assessments that align with caregivers as human evaluators. It is crucial due to the direct relevance of the information provided to the health or condition of the elderly. The motivation behind this proposal stems from challenges observed in previous models, such as BERTScore, which encountered difficulties in effectively evaluating the nurse care record domain, consistently providing quality assessments of generated sentence suggestions above 60%. Additionally, while widely used, cosine similarity exhibits limitations concerning word order, leading to potential misjudgments of semantical differences within similar word sets. Similarly, relying on lexical overlap, ROUGE tends to overlook semantic accuracy. Furthermore, despite its utility, BLEU neglects semantic coherence in its evaluations. EmbedHDP excels in evaluating nurse care records by effectively handling a variety of sentence structures and medical terminology and providing differentiated and contextually relevant assessments. We used a dataset comprising 320 pairs of sentences with correspondingly equivalent lengths. The results revealed that EmbedHDP outperformed other evaluation models, achieving a coefficient score of 61%, followed by cosine similarity with a score of 59%, and BERTScore with 58%. This shows the effectiveness of our proposed approach in improving the evaluation of sentence suggestions in nursing care record applications.

Sentence stimuli pervade psycholinguistics research. Yet limited attention has been paid to the automatic construction of sentence stimuli. Given their linguistic capabilities, this study investigated the efficacy of ChatGPT/AI tools in generating sentence stimuli. In three psycholinguistic experiments, this study examined the acceptability and validity of AI-formulated auditory sentences and written sentences in two languages: English and Arabic. Participants gave English AI-generated stimuli similar to or higher acceptability ratings than human-composed stimuli, but the opposite was observed for Arabic AI-generated stimuli. The validity of AI-developed stimuli relied on the study design, with only Experiment 2 demonstrating the target psycholinguistic effect (L1/L2 morphosyntactic prediction). These results highlight the promising role of AI as a stimuli developer which could facilitate psycholinguistic research and increase its diversity. Implications for psycholinguistic research were discussed.

Colonoscopy is used for colorectal cancer (CRC) screening. Extracting details of the colonoscopy findings from free text in electronic health records (EHRs) can be used to determine patient risk for CRC and colorectal screening strategies. In this study, we developed and evaluated the accuracy of a deep learning model framework to extract information for the clinical decision support system to analyze relevant free-text colonoscopy reports, including indications, pathology, and findings notes. The Bio-Bi-LSTM-CRF framework was developed using Bidirectional Long Short-term Memory (Bi-LSTM) and Conditional Random Fields (CRF) to extract several clinical features from these free-text reports, including indications for the colonoscopy, findings during the colonoscopy, and pathology of the resected material. We then trained the Bio-Bi-LSTM-CRF and existing Bi-LSTM-CRF models on 80% of 4,000 manually annotated notes obtained from the colonoscopy reports of 3,867 patients. The clinical notes were from a group of patients aged over 40 years old enrolled in four Veterans Affairs Medical Centers. A total of 10% of the remaining annotated notes were used to train hyperparameter, while the remaining 10% were used to evaluate the accuracy of our model (Bio-Bi-LSTM-CRF) and to compare the results with the outcomes obtained using Bi-LSTM-CRF. The results of our experiments showed that the bidirectional encoder representations by integrating dictionary function vector from Bio-Bi-LSTM-CRF and strategies character sequence embedding approach is an effective way to identify colonoscopy features from EHR-extracted clinical notes. Therefore, the Bio-Bi-LSTM-CRF model is concluded to be capable of creating new opportunities to identify patients at risk for colon cancer and to study their health outcomes.

We propose that the short−term memory (STM), when processing a sentence, uses two uncorrelated processing units in series. The clues for conjecturing this model emerge from studying many novels of the Italian and English Literatures. This simple model, referring to the surface of language, seems to describe mathematically the input-output characteristics of a complex mental process involved in a reading/writing a sentence. We show that there are no significant mathematical/statistical differences between the two literary corpora by considering deep-language variables and linguistic communication channels, therefore, the surface mathematical structure of alphabetical languages is very deeply rooted in human mind, independently of the language used. The first processing unit is linked to the number of words between two contiguous interpunctions, variable Ip, approximately ranging in Miller’s 7±2 range; the second unit is linked to the number of Ip’s contained in a sentence, variable MF, ranging approximately from 1 to 6. The overall capacity required to process fully of a sentence ranges from 8.3 to 61.23 words, values that can be converted into time by assuming a reading speed, giving the range 2.64~19.54 seconds for fast-reading and 5.3~30.1 seconds for average reader. Since a sentence conveys meaning, the surface features we have found might be the starting point to arrive at an Information Theory that includes meaning.