We interpret the predictions of any blackbox structured input-structured output model around a specific input-output pair. Our method returns an “explanation” consisting of groups of input-output tokens that are causally related. These dependencies are inferred by querying the black-box model with perturbed inputs, generating a graph over tokens from the responses, and solving a partitioning pr...

• We divide the recipes into phases, and use a global structure prediction component to assign different subgenerators generate recipe phases. incorporate attention mechanism get phase-aware features, which are input for produce better Our proposed framework outperforms previous state-of-the-art results. Recipe generation from food images ingredients is challenging task, requires interpretation...

We describe a new system for automatic compiler retargeting. The system combines the advantages of C code generating compilers (fast retargeting) with the advantages of compilers based on speciica-tion driven code generation (fast native code generation). The basic idea is to use the native C compiler to discover architectural features (such as instruction set, register set, addressing modes, i...

The goal of this thesis is to explore how foreign language learning can be facilitated through the use of intelligent computer-assisted language learning (ICALL) based on natural language processing (NLP) methods. ICALL was provided in the form of a task-based dialog system that gives corrective feedback.We investigated how different parameters of the interaction affect the learning progress. B...

Much of NLP tries to map structured input (sentences) to some form of structured output (tag sequences, parse trees, semantic graphs, or translated/paraphrased/compressed sentences). Thus structured prediction and its learning algorithm are of central importance to us NLP researchers. However, when applying machine learning to structured domains, we often face scalability issues for two reasons:

This paper presents an approach to student’s errors diagnosis in intelligent tutoring systems and to the remedial instruction to overcome those errors. Our contribution aris es from two key components. Firstly, a diagnosis model, which based on the nature of the learner’s input, as well as its exercise knowledge model., uses Bayesian induction (a posteriori maximization) to find the most probab...

This essay focuses on metadiscourse, a name for elements of texts that convey meanings other than  those  that  are  primarily  referential.    The  essay  provides  some  theoretical  background  to the  study  of  metadiscourse,  briefly  reviews  a  taxonomy  of  metadiscourse,  and  explores  four reasons why the study of metadiscourse is interesting and important: (a) Such study shows how ...

this study investigates and compares the efficacy of differentiated instruction and traditional-based instruction on enhancing iranian students’ reading comprehension. eight elementary, intermediate, and advanced classrooms from 1 language institute were chosen, and based on their performance on the pretests were further divided into 4 control and 4 treatment groups. flexible grouping, tiered i...