maggicl
/
kse-01
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Added commit id in report

This commit is contained in:
Claudio Maggioni 2023-11-08 22:31:16 +01:00
parent b8e0a2c3c4
commit 7253cf8590
2 changed files with 113 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
report/main.pdf
View File

Binary file not shown.

214
report/main.tex
View File

@ -34,6 +34,7 @@
\usepackage{subcaption}
\usepackage{amssymb}
\usepackage{amsmath}
\usepackage{changepage}
\usepackage{hyperref}
\title{Knowledge Management and Analysis \\ Project 01: Code Search}
@ -42,52 +43,63 @@
\begin{document}
\maketitle
\maketitle
\subsection*{Section 1 - Data Extraction}
\begin{adjustwidth}{-4cm}{-4cm}
\centering
\begin{tabular}{cc}
\toprule
Repository URL & \url{https://github.com/kamclassroom2022/project-01-multi-search-maggicl} \\
Commit ID & \texttt{b8e0a2c3c41249e45b233b55607e0b04ebe1bad0} \\ \bottomrule
\end{tabular}
\end{adjustwidth}
\vspace{1cm}
The data extraction (implemented in the script \texttt{extract-data.py}) process scans through the files in the
TensorFlow project to extract Python docstrings and symbol names for functions, classes and methods. A summary of the
number of features extracted can be found in table~\ref{tab:count1}. The collected figures show that the number of
classes is more than half the number of files, while the number of functions is about twice the number of files.
Additionally, the data shows that a class has slightly more than 2 methods in it on average.
\begin{table}[H]
\centering
\begin{tabular}{cc}
\hline
Type & Number \\
\hline
Python files & 2817 \\
Classes & 1882 \\
Functions & 4565 \\
Methods & 5817 \\
\hline
\end{tabular}
\caption{Count of created classes and properties.}
\label{tab:count1}
\end{table}
\subsection*{Section 1 - Data Extraction}
\subsection*{Section 2: Training of search engines}
The data extraction (implemented in the script \texttt{extract-data.py}) process scans through the files in the
TensorFlow project to extract Python docstrings and symbol names for functions, classes and methods. A summary of the
number of features extracted can be found in table~\ref{tab:count1}. The collected figures show that the number of
classes is more than half the number of files, while the number of functions is about twice the number of files.
Additionally, the data shows that a class has slightly more than 2 methods in it on average.
The training and model execution of the search engines is implemented in the Python script \texttt{search-data.py}.
The training model loads the data extracted by \texttt{extract-data.py} and uses as classification features the
identifier name and only the first line of the comment docstring. All other comment lines are filtered out as this
significantly increases performance when evaluating the models.
\begin{table}[H]
\centering
\begin{tabular}{cc}
\toprule
Type & Number \\
\midrule
Python files & 2817 \\
Classes & 1882 \\
Functions & 4565 \\
Methods & 5817 \\
\bottomrule
\end{tabular}
\caption{Count of created classes and properties.}
\label{tab:count1}
\end{table}
The script is able to search a given natural language query among the extracted TensorFlow corpus using four techniques.
These are namely: Word Frequency Similarity, Term-Frequency Inverse Document-Frequency (TF-IDF) Similarity, Latent
Semantic Indexing (LSI), and Doc2Vec.
\subsection*{Section 2: Training of search engines}
An example output of results generated from the query ``Gather gpu device info'' for the word frequency, TF-IDF, LSI
and Doc2Vec models are shown in
figures~\ref{fig:search-freq},~\ref{fig:search-tfidf},~\ref{fig:search-lsi}~and~\ref{fig:search-doc2vec} respectively.
All four models are able to correctly report the ground truth required by the file \texttt{ground-truth-unique.txt} as
the first result with $>90\%$ similarity, with the except of the Doc2Vec model which reports $71.63\%$ similarity.
The training and model execution of the search engines is implemented in the Python script \texttt{search-data.py}.
The training model loads the data extracted by \texttt{extract-data.py} and uses as classification features the
identifier name and only the first line of the comment docstring. All other comment lines are filtered out as this
significantly increases performance when evaluating the models.
\begin{figure}[b]
\small
\begin{verbatim}
The script is able to search a given natural language query among the extracted TensorFlow corpus using four techniques.
These are namely: Word Frequency Similarity, Term-Frequency Inverse Document-Frequency (TF-IDF) Similarity, Latent
Semantic Indexing (LSI), and Doc2Vec.
An example output of results generated from the query ``Gather gpu device info'' for the word frequency, TF-IDF, LSI
and Doc2Vec models are shown in
figures~\ref{fig:search-freq},~\ref{fig:search-tfidf},~\ref{fig:search-lsi}~and~\ref{fig:search-doc2vec} respectively.
All four models are able to correctly report the ground truth required by the file \texttt{ground-truth-unique.txt} as
the first result with $>90\%$ similarity, with the except of the Doc2Vec model which reports $71.63\%$ similarity.
\begin{figure}[b]
\small
\begin{verbatim}
Similarity: 90.45%
Python function: gather_gpu_devices
Description: Gather gpu device info. Returns: A list of test_log_pb2.GPUInf...
@ -118,13 +130,13 @@ Description: Gather list of devices available to TensorFlow. Returns: A lis...
File: tensorflow/tensorflow/tools/test/system_info_lib.py
Line: 126
\end{verbatim}
\caption{Search result output for the query ``Gather gpu device info'' using the word frequency similarity model.}
\label{fig:search-freq}
\end{figure}
\caption{Search result output for the query ``Gather gpu device info'' using the word frequency similarity model.}
\label{fig:search-freq}
\end{figure}
\begin{figure}[b]
\small
\begin{verbatim}
\begin{figure}[b]
\small
\begin{verbatim}
Similarity: 90.95%
Python function: gather_gpu_devices
Description: Gather gpu device info. Returns: A list of test_log_pb2.GPUInf...
@ -154,13 +166,13 @@ Python function: info
File: tensorflow/tensorflow/python/platform/tf_logging.py
Line: 167
\end{verbatim}
\caption{Search result output for the query ``Gather gpu device info'' using the TF-IDF model.}
\label{fig:search-tfidf}
\end{figure}
\caption{Search result output for the query ``Gather gpu device info'' using the TF-IDF model.}
\label{fig:search-tfidf}
\end{figure}
\begin{figure}[b]
\small
\begin{verbatim}
\begin{figure}[b]
\small
\begin{verbatim}
Similarity: 98.38%
Python function: gather_gpu_devices
Description: Gather gpu device info. Returns: A list of test_log_pb2.GPUInf...
@ -190,13 +202,13 @@ Python method: get_var_on_device
File: tensorflow/tensorflow/python/distribute/packed_distributed_variable.py
Line: 90
\end{verbatim}
\caption{Search result output for the query ``Gather gpu device info'' using the LSI model.}
\label{fig:search-lsi}
\end{figure}
\caption{Search result output for the query ``Gather gpu device info'' using the LSI model.}
\label{fig:search-lsi}
\end{figure}
\begin{figure}[b]
\small
\begin{verbatim}
\begin{figure}[b]
\small
\begin{verbatim}
Similarity: 71.63%
Python function: gather_gpu_devices
Description: Gather gpu device info. Returns: A list of test_log_pb2.GPUInf...
@ -227,59 +239,59 @@ Description: A list of device names for CPU hosts. Returns: A list of devic...
File: tensorflow/tensorflow/python/tpu/tpu_embedding.py
Line: 1011
\end{verbatim}
\caption{Search result output for the query ``Gather gpu device info'' using the Doc2Vec model.}
\label{fig:search-doc2vec}
\end{figure}
\caption{Search result output for the query ``Gather gpu device info'' using the Doc2Vec model.}
\label{fig:search-doc2vec}
\end{figure}
\subsection*{Section 3: Evaluation of search engines}
\subsection*{Section 3: Evaluation of search engines}
The evaluation over the given ground truth to compute precision, recall, and the T-SNE plots is performed by the script
\texttt{prec-recall.py}. The calculated average precision and recall values are reported in table~\ref{tab:tab2}.
The evaluation over the given ground truth to compute precision, recall, and the T-SNE plots is performed by the script
\texttt{prec-recall.py}. The calculated average precision and recall values are reported in table~\ref{tab:tab2}.
Precision and recall are quite high for all models.
The word frequency model has the highest precision and recall ($93.33\%$ and $100.00\%$ respectively), while the Doc2Vec
model has the lowest precision ($73.33\%$) and lowest recall ($80.00\%$).
Precision and recall are quite high for all models.
The word frequency model has the highest precision and recall ($93.33\%$ and $100.00\%$ respectively), while the Doc2Vec
model has the lowest precision ($73.33\%$) and lowest recall ($80.00\%$).
\begin{table}[H]
\centering
\begin{tabular}{ccc}
\hline
Engine & Avg Precision & Recall \\
\hline
Frequencies & 93.33\% & 100.00\% \\
TD-IDF & 90.00\% & 90.00\% \\
LSI & 90.00\% & 90.00\% \\
Doc2Vec & 73.33\% & 80.00\% \\
\hline
\end{tabular}
\caption{Evaluation of search engines.}
\label{tab:tab2}
\end{table}
\begin{table}[H]
\centering
\begin{tabular}{ccc}
\toprule
Engine & Avg Precision & Recall \\
\midrule
Frequencies & 93.33\% & 100.00\% \\
TD-IDF & 90.00\% & 90.00\% \\
LSI & 90.00\% & 90.00\% \\
Doc2Vec & 73.33\% & 80.00\% \\
\bottomrule
\end{tabular}
\caption{Evaluation of search engines.}
\label{tab:tab2}
\end{table}
\subsection*{TBD Section 4: Visualisation of query results}
\subsection*{TBD Section 4: Visualisation of query results}
The two-dimensional T-SNE plots (computed with perplexity $= 2$) for the LSI and Doc2Vec models are respectively in
figures~\ref{fig:tsne-lsi}~and~\ref{fig:tsne-doc2vec}.
The two-dimensional T-SNE plots (computed with perplexity $= 2$) for the LSI and Doc2Vec models are respectively in
figures~\ref{fig:tsne-lsi}~and~\ref{fig:tsne-doc2vec}.
The T-SNE plot for the LSI model shows evidently the presence of outliers in the search result. The Doc2Vec plot shows
fewer outliers and more distinct clusters for the results of each query and the query vector itself. However, even
considering the good performance for both models, it is hard to distinguish from the plots given distinct ``regions''
where results and their respective query are located.
The T-SNE plot for the LSI model shows evidently the presence of outliers in the search result. The Doc2Vec plot shows
fewer outliers and more distinct clusters for the results of each query and the query vector itself. However, even
considering the good performance for both models, it is hard to distinguish from the plots given distinct ``regions''
where results and their respective query are located.
\begin{figure}
\begin{center}
\includegraphics[width=\textwidth]{../out/lsi_plot}
\caption{T-SNE plot for the LSI model over the queries and ground truths given in \texttt{ground-truth-unique.txt}.}
\label{fig:tsne-lsi}
\end{center}
\end{figure}
\begin{figure}
\begin{center}
\includegraphics[width=\textwidth]{../out/lsi_plot}
\caption{T-SNE plot for the LSI model over the queries and ground truths given in \texttt{ground-truth-unique.txt}.}
\label{fig:tsne-lsi}
\end{center}
\end{figure}
\begin{figure}
\begin{center}
\includegraphics[width=\textwidth]{../out/doc2vec_plot}
\caption{T-SNE plot for the Doc2Vec model over the queries and ground truths given in \texttt{ground-truth-unique.txt}.}
\label{fig:tsne-doc2vec}
\end{center}
\end{figure}
\begin{figure}
\begin{center}
\includegraphics[width=\textwidth]{../out/doc2vec_plot}
\caption{T-SNE plot for the Doc2Vec model over the queries and ground truths given in \texttt{ground-truth-unique.txt}.}
\label{fig:tsne-doc2vec}
\end{center}
\end{figure}
\end{document}