Language mapping functions: Improving softmax estimation and word embedding quality

Text Genre Detection Using Doc2Vec Word-embedding Language Model

Language and Information ◽

10.29403/li.23.2.2 ◽

2019 ◽

Vol 23 (2) ◽

pp. 23-43

Author(s):

Dongsung Kim

Keyword(s):

Language Model ◽

Word Embedding ◽

Text Genre

Download Full-text

A Simple Word Embedding Model for Lexical Substitution

10.3115/v1/w15-1501 ◽

2015 ◽

Cited By ~ 13

Author(s):

Oren Melamud ◽

Omer Levy ◽

Ido Dagan

Keyword(s):

Word Embedding ◽

Lexical Substitution

Download Full-text

Word Embedding Based Knowledge Representation with Extracting Relationship Between Scientific Terminologies

Intelligent Automation & Soft Computing ◽

10.31209/2019.100000135 ◽

2019 ◽

pp. -1--1

Author(s):

Mucheol Kim ◽

Junho Kim ◽

Mincheol Shin

Keyword(s):

Knowledge Representation ◽

Word Embedding

Download Full-text

Associations between clinical outcome and tractography based on navigated transcranial magnetic stimulation in patients with language-eloquent brain lesions

Journal of Neurosurgery ◽

10.3171/2018.12.jns182988 ◽

2020 ◽

Vol 132 (4) ◽

pp. 1033-1042 ◽

Cited By ~ 3

Author(s):

Nico Sollmann ◽

Alessia Fratini ◽

Haosu Zhang ◽

Claus Zimmer ◽

Bernhard Meyer ◽

...

Keyword(s):

Transcranial Magnetic Stimulation ◽

Brain Tumors ◽

Risk Stratification ◽

Magnetic Stimulation ◽

Diffusion Tensor ◽

Tumor Resection ◽

Navigated Transcranial Magnetic Stimulation ◽

Language Mapping ◽

Significant Difference ◽

Related Pathway

OBJECTIVENavigated transcranial magnetic stimulation (nTMS) in combination with diffusion tensor imaging fiber tracking (DTI FT) is increasingly used to locate subcortical language-related pathways. The aim of this study was to establish nTMS-based DTI FT for preoperative risk stratification by evaluating associations between lesion-to-tract distances (LTDs) and aphasia and by determining a cut-off LTD value to prevent surgery-related permanent aphasia.METHODSFifty patients with left-hemispheric, language-eloquent brain tumors underwent preoperative nTMS language mapping and nTMS-based DTI FT, followed by tumor resection. nTMS-based DTI FT was performed with a predefined fractional anisotropy (FA) of 0.10, 0.15, 50% of the individual FA threshold (FAT), and 75% FAT (minimum fiber length [FL]: 100 mm). The arcuate fascicle (AF), superior longitudinal fascicle (SLF), inferior longitudinal fascicle (ILF), uncinate fascicle (UC), and frontooccipital fascicle (FoF) were identified in nTMS-based tractography, and minimum LTDs were measured between the lesion and the AF and between the lesion and the closest other subcortical language-related pathway (SLF, ILF, UC, or FoF). LTDs were then associated with the level of aphasia (no/transient or permanent surgery-related aphasia, according to follow-up examinations).RESULTSA significant difference in LTDs was observed between patients with no or only surgery-related transient impairment and those who developed surgery-related permanent aphasia with regard to the AF (FA = 0.10, p = 0.0321; FA = 0.15, p = 0.0143; FA = 50% FAT, p = 0.0106) as well as the closest other subcortical language-related pathway (FA = 0.10, p = 0.0182; FA = 0.15, p = 0.0200; FA = 50% FAT, p = 0.0077). Patients with surgery-related permanent aphasia showed the lowest LTDs in relation to these tracts. Thus, LTDs of ≥ 8 mm (AF) and ≥ 11 mm (SLF, ILF, UC, or FoF) were determined as cut-off values for surgery-related permanent aphasia.CONCLUSIONSnTMS-based DTI FT of subcortical language-related pathways seems suitable for risk stratification and prediction in patients suffering from language-eloquent brain tumors. Thus, the current role of nTMS-based DTI FT might be expanded, going beyond the level of being a mere tool for surgical planning and resection guidance.

Download Full-text

Key phrase Extraction by Improving TextRank with an Integration of Word Embedding and Syntactic Information

Recent Advances in Computer Science and Communications ◽

10.2174/2666255813999200820155846 ◽

2020 ◽

Vol 13 ◽

Author(s):

Sheng Zhang ◽

Qi Luo ◽

Yukun Feng ◽

Ke Ding ◽

Daniela Gifu ◽

...

Keyword(s):

Semantic Information ◽

Performance Enhancement ◽

Word Embedding ◽

The Other ◽

Test Set ◽

Pagerank Algorithm ◽

Phrase Extraction ◽

Extraction Algorithm ◽

Syntactic Information ◽

Key Phrase Extraction

Background: As a known key phrase extraction algorithm, TextRank is an analogue of PageRank algorithm, which relied heavily on the statistics of term frequency in the manner of co-occurrence analysis. Objective: The frequency-based characteristic made it a neck-bottle for performance enhancement, and various improved TextRank algorithms were proposed in the recent years. Most of improvements incorporated semantic information into key phrase extraction algorithm and achieved improvement. Method: In this research, taking both syntactic and semantic information into consideration, we integrated syntactic tree algorithm and word embedding and put forward an algorithm of Word Embedding and Syntactic Information Algorithm (WESIA), which improved the accuracy of the TextRank algorithm. Results: By applying our method on a self-made test set and a public test set, the result implied that the proposed unsupervised key phrase extraction algorithm outperformed the other algorithms to some extent.

Download Full-text

A Novel Language Paradigm for Intraoperative Language Mapping: Feasibility and Evaluation

Journal of Clinical Medicine ◽

10.3390/jcm10040655 ◽

2021 ◽

Vol 10 (4) ◽

pp. 655

Author(s):

Katharina Rosengarth ◽

Delin Pai ◽

Frank Dodoo-Schittko ◽

Katharina Hense ◽

Teele Tamm ◽

...

Keyword(s):

Healthy Subjects ◽

Intraoperative Complications ◽

Evaluation Study ◽

Awake Craniotomy ◽

Language Mapping ◽

Tumor Patients ◽

Language Functions ◽

Before And After ◽

Assessment Time ◽

Neuropsychological Outcomes

(1) Background—Mapping language using direct cortical stimulation (DCS) during an awake craniotomy is difficult without using more than one language paradigm that particularly follows the demand of DCS by not exceeding the assessment time of 4 s to prevent intraoperative complications. We designed an intraoperative language paradigm by combining classical picture naming and verb generation, which safely engaged highly relevant language functions. (2) Methods—An evaluation study investigated whether a single trial of the language task could be performed in less than 4 s in 30 healthy subjects and whether the suggested language paradigm sufficiently pictured the cortical language network using functional magnetic resonance imaging (fMRI) in 12 healthy subjects. In a feasibility study, 24 brain tumor patients conducted the language task during an awake craniotomy. The patients’ neuropsychological outcomes were monitored before and after surgery. (3) Results—The fMRI results in healthy subjects showed activations in a language-associated network around the (left) sylvian fissure. Single language trials could be performed within 4 s. Intraoperatively, all tumor patients showed DCS-induced language errors while conducting the novel language task. Postoperatively, mild neuropsychological impairments appeared compared to the presurgical assessment. (4) Conclusions—These data support the use of a novel language paradigm that safely monitors highly relevant language functions intraoperatively, which can consequently minimize negative postoperative neuropsychological outcomes.

Download Full-text