Aiming at the problems of inadequate semantic understanding, improper summary sentences and inaccurate summary in the field of Natural Language Processing (NLP) abstractive automatic summarization, a new automatic summary solution was proposed, including an improved word vector generation technique and an abstractive automatic summarization model. The improved word vector generation technology was based on the word vector generated by the skip-gram method. Combining with the characteristics of abstract, three word features including part of speech, word frequency and inverse text frequency were introduced, which effectively improved the understanding of words. The proposed Bi-MulRnn+ abstractive automatic summarization model was based on sequence-to-sequence (seq2seq) framework and self-encoder structure. By introducing attention mechanism, Gated Recurrent Unit (GRU) gate structure, Bi-directional Recurrent Neural Network (BiRnn) and Multi-layer Recurrent Neural Network (MultiRnn), the model improved the summary accuracy and sentence fluency of abstractive summarization. The experimental results of Large-Scale Chinese Short Text Summarization (LCSTS) dataset show that the proposed scheme can effectively solve the problem of abstractive summarization of short text, and has good performance in Rouge standard evaluation system, improving summary accuracy and sentence fluency.

Focused on the issue that the location fingerprint matching localization algorithm has a large workload of offline database collection in an indoor environment, a fast virtual grid matching algorithm based on Pearson correlation coefficient was proposed. Firstly, the Received Signal Strength Indicator (RSSI) was preprocessed with Gaussian filter to obtain the received signal strength vector. Then, the Bounding-Box method was used to determine the initial virtual grid region. The grid region was rapidly and iteratively subdivided, the distance log vectors of the grid center point to beacon nodes were calculated, and the Pearson correlation coefficients between the received signal strength vector and the distance log vectors were calculated. Finally, the k nearest neighbor coordinates whose correlation coefficients close to -1 were selected, and the optimal estimation position of the undetermined node was determined by the weighted estimation of correlation coefficients. The simulation results show that the localization error of the proposed algorithm is less than 2m in 94.2% probability under the condition of 1m virtual grid and RSSI noise standard deviation of 3dBm. The positioning accuracy is better than that of the location fingerprint matching algorithm, and the RSSI fingerprint database is no longer needed, which greatly reduces the workload of localization.