Official Document: https://1drv.ms/b/c/c91ef16cd60b809e/EZ6AC9Zs8R4ggMmNiwAAAAABa-Su2Cwlg6N9bias7ErzSw?e=x5fV2s

Evolution of technology perception and industries' adoption

How could machine learning and AI impact organizations? This question is particularly pertinent in modern business management research, as it delves into a critical aspect of organizational dynamics in the face of advancing technologies.

I've chosen to focus on the impact of machine learning and AI tools because they've already begun to reshape how work is organized within organizations, whether through the automation of routine tasks, augmentation of decision-making processes, or the emergence of entirely new roles and responsibilities centered around AI integration.

Vader sentiment Analysis Heuristics

The sentiment analysis results were statistically analyzed using Scipy and Scikit-Learn libraries. This involved calculating average sentiment scores across different topics and keyword clusters, allowing for the identification of trends in how AI research is perceived across various organizational contexts. (Scikit-learn: Machine Learning in Python, Pedregosa et al. 2011) The statistical insights were supplemented with qualitative observations derived from the manual extraction of key concepts, adding depth to the interpretation of the data.

Insights into Research Themes and Organizational Perception

The bar chart showed on Figure 3 illustrates the sentiment score of research related to AI and machine learning from 1985 to 2023. The y-axis represents the average sentiment, with values ranging from 0 to 0.7, while the x-axis shows the years. Each bar corresponds to a specific year, with the sentiment score labeled at the top of each bar.

This distribution reinforces the positive perception identified earlier, where AI and machine learning research, particularly in high-growth sectors like technology, transportation, and science, has increasingly been seen as favorable. (Curry et al. 2021). The clustering of sentiment around higher values suggests that organizations view these technologies as beneficial to operational efficiency, innovation, and strategic planning. The significant frequency of neutral sentiment scores aligns with the scientific rigor and objectivity often found in research documents, where results are presented with caution and without emotional bias.

Note

For the Part of Speech (POS) it was used only the following categories proper nouns (NNP, NNPS), technical terms (domain-specific), nouns (NN, NNS), and verbs (VB, VBD, VBG, VBN, VBP, VBZ)

Topics modeling and Table

The table and topic modeling presents a series of topics, each represented by the top keywords that define its thematic focus. These topics are likely the result of a topic modeling analysis, which identified patterns and recurring themes across a collection of AI and machine learning research documents. The terms within each topic provide insights into the specific areas of focus in the research corpus. For example, Topic 1, which includes terms like "economic," "european," and "political," suggests a focus on AI's impact within economic and policy-related contexts, while Topic 3, with terms like "research," "ai," and "design," centers on core AI methodologies and research practices. Other topics, such as Topic 7, with terms like "energy" and "security," indicate AI's application in areas like energy systems and cybersecurity.

Topic 0	Topic 1	Topic 2	Topic 3	Topic 4	Topic 5	Topic 6	Topic 7
search	economic	language	research	system	early	model	energy
algorithm	european	logic	ai	model	surface	method	security
linear	political	information	design	information	showed	performance	channel
case	environmental	reasoning	development	design	late	neural	system
tree	impact	representation	digital	order	temperature	algorithm	connection
solution	article	framework	social	control	observed	network	edge
random	market	theory	work	agent	production	machine	wireless
ai	french	semantic	article	user	material	deep	performance
optimal	financial	heidelberg	human	environment	site	optimization	network
finite	role	fuzzy	management	complex	atlas	training	detection
classical	risk	natural	product	framework	age	efficient	timing
distribution	local	order	machine	dynamic	middle	control	communication

Figure 9 describes the yearly fluctuation in average sentiment across various research topics. The y-axis represents average sentiment scores, while the x-axis shows the timeline from 1985 to 2020. Each colored line corresponds to a different research topic. Throughout the timeline, sentiment across all topics fluctuates, with notable variability in the earlier years, particularly between 1985 and 2000, where sharp rises and falls are observed. However, from 2000 onwards, the trends stabilize, and sentiment shows a gradual upward trend across most topics, particularly from 2015 onward.

The Correlation Matrix in Figure 11 visualizes the relationships between sentiment scores across eight distinct AI research topics. The matrix is color-coded, with darker red shades indicating higher positive correlations and lighter shades reflecting weaker or negative correlations. Strong positive correlations, such as between Topics 1 and 7 (0.95), or Topics 0 and 3 (0.90), suggest thematic or perceptual overlap, meaning that research in these areas tends to evoke similar sentiment reactions. In contrast, weaker correlations, such as between Topic 4 and Topic 2 (0.47), indicate that sentiment perceptions in these areas diverge, possibly due to differences in focus or complexity.

Data

• Research about AI & business in France

  • Articles, research, books, and more founded on Scopus advanced searching, building a Dataset with all the metadata of different sources

  • URL: https://www.scopus.com/standard/marketing.uri

    ( TITLE-ABS-KEY ( ai ) OR ALL ( ai ) AND ALL ( business ) OR ALL ( organization ) OR ALL ( enterprise ) OR ALL ( work ) OR ALL ( profession ) OR ALL ( career ) OR ALL ( affair ) OR ALL ( occupation ) OR ALL ( report ) OR ALL ( market ) OR ALL ( invest ) OR ALL ( trade ) OR ALL ( industry ) OR ALL ( company ) OR ALL ( commerce ) OR ALL ( dealing ) OR ALL ( firm ) OR ALL ( purchase ) OR ALL ( survey ) OR ALL ( manager ) OR ALL ( manage ) OR ALL ( decision ) OR ALL ( digital ) OR ALL ( task ) OR ALL ( automate ) OR ALL ( processes ) OR ALL ( production ) OR ALL ( bot ) OR ALL ( data ) OR ALL ( selling ) OR ALL ( marketing ) OR ALL ( logistics ) OR ALL ( finance ) OR ALL ( fabricate ) OR ALL ( price ) OR ALL ( stock ) OR ALL ( network ) OR ALL ( resource ) OR ALL ( money ) OR ALL ( cash ) OR ALL ( credit ) OR ALL ( institution ) OR ALL ( smart ) OR ALL ( tech ) OR ALL ( case ) OR ALL ( trading ) OR ALL ( area ) OR ALL ( system ) AND NOT ALL ( medicine ) AND NOT ALL ( patients ) AND NOT ALL ( patients ) AND NOT ALL ( health ) AND NOT ALL ( disease ) AND NOT ALL ( covid-19 ) AND NOT ALL ( healthcare ) AND NOT ALL ( viral ) ) AND PUBYEAR > 1969 AND PUBYEAR < 2024 AND ( LIMIT-TO ( AFFILCOUNTRY , "France" ) ) AND ( LIMIT-TO ( SUBJAREA , "SOCI" ) OR LIMIT-TO ( SUBJAREA , "BUSI" ) OR LIMIT-TO ( SUBJAREA , "DECI" ) OR LIMIT-TO ( SUBJAREA , "MULT" ) OR LIMIT-TO ( SUBJAREA , "ECON" ) OR LIMIT-TO ( SUBJAREA , "PSYC" ) OR LIMIT-TO ( SUBJAREA , "ENGI" ) OR LIMIT-TO ( SUBJAREA , "MATH" ) OR LIMIT-TO ( SUBJAREA , "NEUR" ) OR LIMIT-TO ( SUBJAREA , "ENER" ) OR LIMIT-TO ( SUBJAREA , "ARTS" ) ) AND ( LIMIT-TO ( LANGUAGE , "English" ) )
    

• French Government AI related economic information

  • Cases, projects, regulations, and more related to AI in France, the source is the official site of the government, and the dataset created was web scrapped
  • URL: https://www.economie.gouv.fr/recherche-resultat?search_api_views_fulltext=IA&page=0
  • https://www.economie.gouv.fr/

• News of AI-related business, organization, economics and finance

  • The dataset created was web scrapped, applying different queries of searching into google search web scrapping tool defined on the folder "Websccrapping", results are compiled on a single dataset that contains all organizations and business AI-related results

• Enterprises_AI

  • Information about capital, composition, operations, and more about enterprises that lead the use and application of AI
  • URL: https://data.world/aurielle/the-essential-landscape-of-enterprise-a-i-companies/workspace/file?filename=EnterpriseAI.csv

• AI Tools

  • Tools based on different categories to automatize processes and optimize business, there are two sources, one official of French gouvernment datasets and another extracted from Kagle, in which it can be seen tools description and some metadata
  • URL GouvFR: https://www.data.gouv.fr/fr/datasets/4000-outils-ai/
  • URL Kagle: https://www.kaggle.com/datasets/yasirabdaali/740-ai-tools-for-everyone

• Data Scientist vs size of Datasets

  • A group of 100 data scientists from France were interviewed between January 2016 and August 2016 to analyze the potential relationship between hardware and data set sizes. However, it is important to note that the sample size may not represent the entire population. Dataset obtained from Kagle
  • URL Kagle: https://www.kaggle.com/datasets/laurae2/data-scientists-vs-size-of-datasets/data

• Financial impact on CAC40

  • According to the results obtained from previous experiments on this project and guided by topic modeling structure, it has been selected specific companies that are related to AI use cases on their internal processes to analyze how the impact was received. The dataset built based on financial statements and open financial information
  • URL: https://fr.finance.yahoo.com/

Warning

It is necessary to install the libraries listed below and understand the requirements to make the hypothesis and Data Treatment

Principal processes and requirements (AZURE, web scrapping, libraries, etc)

• Topic modeling documentation: https://github.com/piskvorky/gensim?tab=readme-ov-file

• Topic Modeling Gensim quick guide: https://radimrehurek.com/gensim/auto_examples/core/run_core_concepts.html

• Creating Azure resource text analysis on Language Studio to obtain API key and endpoint for processing text analysis

  Sample: https://github.com/Azure/azure-sdk-for-python/blob/main/sdk/textanalytics/azure-ai-textanalytics/samples/sample_analyze_sentiment.py

• Creating virtual environment variables with Azure "text analysis" key and endpoint on a file .env and calling it on Python script as the following structure:

  • Load environment variables from the .env file, these would be the keys and endpoint for used API's - load_dotenv()

  • Access the environment variables

      endpoint = os.getenv("AZURE_TEXT_ANALYSIS_ENDPOINT")
      key = os.getenv("AZURE_TEXT_ANALYSIS_KEY")
    

• Microsoft Visual C++ 14.0 or greater is required. Get it with "Microsoft C++ Build Tools": https://visualstudio.microsoft.com/visual-cpp-build-tools/

  Dependence list: ['pybind11<2.6.2', 'psutil', "numpy>=1.10.0,<1.17 ; python_version=='2.7'", "numpy>=1.10.0 ; python_version>='3.5'"]     
  running bdist_wheel
  running build
  running build_ext
  Extra compilation arguments: ['/EHsc', '/openmp', '/O2', '/DVERSION_INFO=\\"2.1.1\\"']    
  building 'nmslib' extension
  

• Required python Libraries

  For Webscrapping

    pip install matplotlib
    pip install wordcloud
    pip install bs4
    pip install selenium
    pip install webdriver_manager
    pip install undetected_chromedriver
    pip install requests beautifulsoup4
    pip install googlesearch-python
    pip install streamlit
  

  For Topic Modeling

    pip install --upgrade gensim
    pip install Pyro4
    pip install Sphinx
    pip install annoy
    pip install memory-profiler
    pip install nltk
    pip install nmslib (previous requirement of C++ development tools)
    pip install POT
    pip install scikit-learn
    pip install sphinx-gallery
    pip install sphinxcontrib-napoleon
    pip install sphinxcontrib-programoutput
    pip install statsmodels
    pip install testfixtures
    pip install spacy # For stop words
    python -m spacy download en_core_web_sm  # for English
    python -m spacy download fr_core_news_sm $ for French
    pip install gensim nltk pyLDAvis
    pip install gensim nltk matplotlib
    pip install Flask
    pip install TextBlob
  

  For Azure and other API's connection

    pip install azure-ai-textanalytics
    pip install azure-identity
    pip install python-dotenv
  

  For BERT transformers with pytorch

    pip install torch
    pip install transformers torch
  

  For better optimization with NumPy and OpenBLAS on the NLP with topic modeling

    pip install numpy --only-binary :numpy: numpy
    pip install numpy --no-binary numpy
  

Note

References: disposed on the file 'references.txt'