python - 如何在 Linux 系统上使用 SyntaxNet 输出来操作执行命令，例如将文件保存在文件夹中

Question

已下载并训练SyntaxNet ，我正在尝试编写一个程序，可以打开新的/现有的文件，例如 AutoCAD 文件，并通过分析文本将文件保存在特定目录中:打开 LibreOffice 文件 X。将 SyntaxNet 的输出考虑为:

echo "save AUTOCAD file  X in directory Y" | ./test.sh > output.txt


Input: save AUTOCAD file X in directory Y
Parse:
save VB ROOT
 +-- X NNP dobj
 |   +-- file NN compound
 |       +-- AUTOCAD CD nummod
 +-- directory NN nmod
     +-- in IN case
     +-- Y CD nummod

首先，我考虑将已解析的文本更改为 XML 格式，然后使用语义分析(如 SPARQL)解析 XML 文件以找到 ROOT=save、dobj=X 和 nummode=Y 并编写一个 python 程序可以执行和文中说的一样

1. 我不知道如果我将解析的文本更改为 XML，然后使用使用查询的语义分析来匹配 ROOT与其对应的函数或脚本保存 dobj ，在 nummode 中提到的目录中

2. 我有一些想法可以使用 subprocess 将 python 连接到终端包，但我没有找到任何可以帮助我从终端保存例如 AUTOCAD 文件或任何其他文件的东西，或者我是否需要编写脚本，.sh , 在 python 的帮助下？

我对文本的句法和语义分析进行了大量研究，例如 Christian Chiarcos, 2011 , Hunter and Cohen, 2006和 Verspoor et al., 2015 ，还研究了Microsoft Cortana , Sirius , google now但他们都没有详细说明如何将已解析的文本更改为执行命令，这让我得出结论，这项工作太容易谈论了，但是因为我不是计算机科学专业的学生，​​所以我不知道我能做些什么。

Answer 1

我是 Computer Science World 和 SyntaxNet 的初学者。我写了一个简单的 SyntaxNet-Python 算法，它使用 SyntaxNet 来分析用户插入的文本命令，“打开我用 LibreOffice writer 和 laboratory writer 写的文件簿”，然后用 python 算法分析 SyntaxNet 输出，以便转它是一个执行命令，在这种情况下打开一个文件，具有任何支持的格式，在 Linux，Ubuntu 14.04) 环境中使用 LibreOffice。你可以看到here LibreOffice 定义的不同命令行，以便在此包中使用不同的应用程序。

1. 安装并运行SyntaxNet后(安装过程在here中解释)，shell脚本被打开demo.sh ~/models/syntaxnet/suntaxnet/ 目录和 conl2tree 函数(第 54 到 56 行)被删除以获得 >制表符分隔从 SyntaxNet 输出而不是树格式输出。

2. 在终端窗口中输入此命令:

   echo '打开我用libreOffice writer 和labre writer 写的文件本' | syntaxnet/demo.sh > 输出.txt

output.txt文件保存在demo.sh所在目录下，如下图:

1. 将 output.txt 作为输入文件并使用以下 python 算法分析 SyntaxNet 输出并从 LibreOffice 包中识别您想要的目标应用程序的文件名和用户想要的命令使用。

#!/bin/sh

import csv

import subprocess

import sys

import os

#get SyntaxNet output as the Python algorithm input file
filename='/home/username/models/syntaxnet/work/output.txt'

#all possible executive commands for opening any file with any format with Libreoffice file
commands={
('open',  'libreoffice',  'writer'):  ('libreoffice', '--writer'),
('open',  'libreoffice',  'calculator'):  ('libreoffice' ,'--calc'),
('open',  'libreoffice',  'draw'):  ('libreoffice' ,'--draw'),
('open',  'libreoffice',  'impress'): ('libreoffice' ,'--impress'),
('open',  'libreoffice',  'math'):  ('libreoffice' ,'--math'),
('open',  'libreoffice',  'global'):  ('libreoffice' ,'--global'),
('open',  'libreoffice',  'web'): ('libreoffice' ,'--web'),
('open',  'libreoffice',  'show'):  ('libreoffice', '--show'),
}
#all of the possible synonyms of the application from Libreoffice 
comments={
 'writer': ['word','text','writer'],
 'calculator': ['excel','calc','calculator'],
 'draw': ['paint','draw','drawing'],
 'impress': ['powerpoint','impress'],
 'math': ['mathematic','calculator','math'],
 'global': ['global'],
 'web': ['html','web'],
 'show':['presentation','show']
 }

root ='ROOT'            #ROOT of the senctence
noun='NOUN'             #noun tagger
verb='VERB'             #verb tagger
adjmod='amod'           #adjective modifier
dirobj='dobj'           #direct objective
apposmod='appos'        # appositional modifier
prepos_obj='pobj'       # prepositional objective
app='libreoffice'       # name of the package
preposition='prep'      # preposition
noun_modi='nn'          # noun modifier 

#read from Syntaxnet output tab delimited textfile
def readata(filename):
    file=open(filename,'r')
    lines=file.readlines()
    lines=lines[:-1]
    data=csv.reader(lines,delimiter='\t')
    lol=list(data)
    return  lol

# identifies the action, the name of the file and whether the user mentioned the name of the application implicitely  
def exe(root,noun,verb,adjmod,dirobj,apposmod,commands,noun_modi):
    interprete='null'
    lists=readata(filename)
    for sublist in lists:
        if sublist[7]==root and sublist[3]==verb: # when the ROOT is verb the dobj is probably the name of the file you want to have
                action=sublist[1]
                dep_num=sublist[0]
                for sublist in lists:
                    if sublist[6]==dep_num and sublist[7]==dirobj:
                        direct_object=sublist[1]
                        dep_num=sublist[0]
                        dep_num_obj=sublist[0]
                        for sublist in lists:
                            if direct_object=='file' and sublist[6]==dep_num_obj and sublist[7]==apposmod:
                                direct_object=sublist[1]
                            elif  direct_object=='file' and sublist[6]==dep_num_obj and sublist[7]==adjmod:
                                direct_object=sublist[1]
                for sublist in lists:
                    if sublist[6]==dep_num_obj and sublist[7]==adjmod:
                            for key, v in  comments.iteritems():
                                if sublist[1] in v:
                                    interprete=key
                for sublist in lists:
                    if sublist[6]==dep_num_obj and sublist[7]==noun_modi:
                        dep_num_nn=sublist[0]
                        for key, v in  comments.iteritems():
                            if sublist[1] in v:
                                interprete=key
                                print interprete
                        if interprete=='null':
                            for sublist in lists:
                                if sublist[6]==dep_num_nn and sublist[7]==noun_modi:
                                    for key, v in  comments.iteritems():
                                        if sublist[1] in v:
                                            interprete=key
        elif  sublist[7]==root and sublist[3]==noun: # you have to find the word which is in a adjective form and depends on the root
            dep_num=sublist[0]
            dep_num_obj=sublist[0]
            direct_object=sublist[1]
            for sublist in lists:
                if sublist[6]==dep_num and sublist[7]==adjmod:
                    actionis=any(t1==sublist[1] for (t1, t2, t3) in commands)
                    if actionis==True:
                        action=sublist[1]
                elif sublist[6]==dep_num and sublist[7]==noun_modi:
                    dep_num=sublist[0]
                    for sublist in lists:
                        if sublist[6]==dep_num and sublist[7]==adjmod:
                            if any(t1==sublist[1] for (t1, t2, t3) in commands):
                                action=sublist[1]
            for sublist in lists:
                if direct_object=='file' and sublist[6]==dep_num_obj and sublist[7]==apposmod and sublist[1]!=action:
                    direct_object=sublist[1]
                if  direct_object=='file' and sublist[6]==dep_num_obj and sublist[7]==adjmod and sublist[1]!=action:
                    direct_object=sublist[1]
            for sublist in lists:
                if sublist[6]==dep_num_obj and sublist[7]==noun_modi:
                    dep_num_obj=sublist[0]
                    for key, v in  comments.iteritems():
                        if sublist[1] in v:
                            interprete=key
                        else:
                            for sublist in lists:
                                if sublist[6]==dep_num_obj and sublist[7]==noun_modi:
                                    for key, v in  comments.iteritems():
                                        if sublist[1] in v:
                                            interprete=key
    return action, direct_object, interprete

action, direct_object, interprete = exe(root,noun,verb,adjmod,dirobj,apposmod,commands,noun_modi)

# find the application (we assume we know user want to use libreoffice but we donot know what subapplication should be used)
def application(app,prepos_obj,preposition,noun_modi):
    lists=readata(filename)
    subapp='not mentioned'
    for sublist in lists:
        if sublist[1]==app:
            dep_num=sublist[6]
            for sublist in lists:
                if sublist[0]==dep_num and sublist[7]==prepos_obj:
                    actioni=any(t3==sublist[1] for (t1, t2, t3) in commands)
                        if actioni==True:
                            subapp=sublist[1]
                        else:
                            for sublist in lists:
                                if sublist[6]==dep_num and sublist[7]==noun_modi:
                                    actioni=any(t3==sublist[1] for (t1, t2, t3) in commands)
                                    if actioni==True:
                                        subapp=sublist[1]
                        elif sublist[0]==dep_num and sublist[7]==preposition:
                            sublist[6]=dep_num
                            for subline in lists:
                                if subline[0]==dep_num and subline[7]==prepos_obj:
                                    if any(t3==sublist[1] for (t1, t2, t3) in commands):
                                        subapp=sublist[1]
                                    else:
                                        for subline in lists:
                                            if subline[0]==dep_num and subline[7]==noun_modi:
                                                if any(t3==sublist[1] for (t1, t2, t3) in commands):
                                                    subapp=sublist[1]
    return subapp

sub_application=application(app,prepos_obj,preposition,noun_modi)

if sub_application=='not mentioned' and interprete!='null':
    sub_application=interprete
elif sub_application=='not mentioned' and interprete=='null':
    sub_application=interprete

# the format of file
def format_function(sub_application):
    subapp=sub_application
    Dobj=exe(root,noun,verb,adjmod,dirobj,apposmod,commands,noun_modi)[1]
    if subapp!='null':
        if subapp=='writer':
            a='.odt'
            Dobj=Dobj+a
        elif subapp=='calculator':
            a='.ods'
            Dobj=Dobj+a
        elif subapp=='impress':
            a='.odp'
            Dobj=Dobj+a
        elif subapp=='draw':
            a='.odg'
            Dobj=Dobj+a
        elif subapp=='math':
            a='.odf'
            Dobj=Dobj+a
        elif subapp=='math':
            a='.odf'
            Dobj=Dobj+a
        elif subapp=='web':
            a='.html'
            Dobj=Dobj+a
    else:
        Dobj='null'
    return Dobj

def get_filepaths(directory):
    myfile=format_function(sub_application)
    file_paths = []  # List which will store all of the full filepaths.
    # Walk the tree.
    for root, directories, files in os.walk(directory):
        for filename in files:
        # Join the two strings in order to form the full filepath.
            if filename==myfile:
                filepath = os.path.join(root, filename)
                file_paths.append(filepath)  # Add it to the list.
    return file_paths  # Self-explanatory.

# Run the above function and store its results in a variable.
full_file_paths = get_filepaths("/home/ubuntu/")

if full_file_paths==[]:
    print 'No file with name %s is found' % format_function(sub_application)
if full_file_paths!=[]:
    path=full_file_paths
    prompt='> '
    if len(full_file_paths) >1:
        print full_file_paths
        print 'which %s do you mean?'% subapp
        inputname=raw_input(prompt)
        if inputname in full_file_paths:
            path=inputname
        #the main code structure
    if sub_application!='null':
        command= commands[action,app,sub_application]
        subprocess.call([command[0],command[1],path[0]])
    else:
        print "The sub application is not mentioned clearly"

我再次声明我是初学者，代码可能看起来不那么整洁或专业，但我只是尝试使用我对这个迷人的所有知识SyntaxNet 一个实用的算法。这个简单的算法可以打开文件:

1. LibreOffice 支持的任何格式，例如.odt,.odf,.ods,.html,.odp.

2. 可以理解LibreOffice中不同应用的隐式引用，例如:“用libreoffice打开文本文件簿”而不是“用libreoffice writer打开文件簿”

3. 可以克服 SyntaxNet 解释被称为形容词的文件名的问题。