e-tissage

Why is a graph structure, the best way to organize complex data ?

Véronique Gendner — Thu, 10 Apr 2025 15:27:25 +0000

Visually analysing data with a graph structure and the yWorks Project Jupyter module

Véronique Gendner — Mon, 16 Dec 2024 15:33:46 +0000

The yFiles Graphs for Jupyter widget exists for different graph technologies : Networkx, igraph, neo4j, pygraphviz and any structured Python dictionaries and lists https://www.yworks.com/products/yfiles-graphs-for-jupyter

My Cypher Query Takes Too Long. What can I do ?

Véronique Gendner — Wed, 06 Nov 2024 18:21:43 +0000

The purpose of this talk given at NODES 2024 is to

give general understanding of how the cypher query planner works and the kind of things you need to know to analyses what takes times in a query processing
give some insight on how to use EXPLAIN and PROFILE to identify what takes time in running a specific query
give examples for you to build a better understanding of how cypher works under the hood

Fine-tuning for real time constrains or other very demanding performance constrains will NOT be addressed. This talk is for general understanding of difficulties to traverse graphs, hints about how the query planner plans query processing and some guidelines about the different parameters query processing time depends upon.

Slides with clickable links

Download slides with clickable links

Slides with written comments

Graphs & LLM : a visual introduction

Véronique Gendner — Thu, 12 Sep 2024 14:54:44 +0000

Explore Core Concepts – Gain a foundational understanding of how graph technologies enhance Generative AI applications. Learn about essential terms and concepts such as embeddings, vector search, and Graph #RAG.
Learn about the #Neo4j property graph model – Understand how it can be leveraged to structure and query data effectively.
Understand Limitations and Opportunities – Identify some of the current limitations of integrating graphs with LLMs and explore opportunities for overcoming these challenges.
Receive Practical Resources – Walk away with a cheat sheet that includes technical basics, helpful documentation, and essential resources. You’ll also receive code examples for Neo4j and #LangChain to help you get started.

Whether you’re new to the potential of combining graphs with LLMs or need to convey these ideas to non-technical colleagues, this talk will help you move forward!

Slides with clickable links

Download slides with clickable links

Slides with written comments

Introduction To Cypher

Véronique Gendner — Wed, 27 Mar 2024 11:25:06 +0000

Slides & hands-on curriculum of the Introduction to Cypher presented in Malmö, Sweden for Pink Programming

Hands-On

Introduction to Graph query language Cypher from Véronique Gendner

Knowledge Graph Conference demo session : Neodash

Véronique Gendner — Mon, 22 May 2023 15:12:23 +0000

Graphs are a very useful formal structure because you can represent in one place, the precise complexity of your data, by interrelating lists, hierarchies, categorizations and tables. You can also avoid being lost in complexity by choosing what to expose for a specific purpose, in any of those more basic formal structures.

Here is a 5 minutes introduction to NeoDash, a visualisation tool that allows to easily display lists, tables, graphs, charts & maps, extracted from a Neo4j database. It was presented in the demo session of the Knowledge Graph Conference.

NODES 2022 : Genealogy with different graph technologies for data collection & visualization

Véronique Gendner — Thu, 17 Nov 2022 14:09:56 +0000

Graphs are a very interesting structure to get the best of human thinking combined with automatic processing. Through a very intuitively graph based use case – genealogy – the following presentation at the Nodes 2022 conference takes genealogical researches as an example to illustrate possibilities that can be applied to many domain of research. It is demoing several graph issues and technologies, from data collection to visualization, including data transfer from one app to another:

documentation collection, in a graph structure, with TheBrain Technologies application,
Gramps visualizations based on a family tree specific layout algorithm
Bloom for visual and interactive exploration of data structured in graph and
yWorks to make precisely selected information of interest, stand out on a graph display
as well as Neo4j graph database as a pivot technology, to make use of graph topology and graph pattern matching

Data collection and export/import process from TheBrain to Gramps via Neo4j are presented, graph models are discussed.

TheBrain is a great User Interface to collect scraps of information as they show up in no specific order and easily relate them in a meaningful way, that is in a graph structure.
It’s great for data collection, but global visualization of the graph is not appropriate, for example to visualize remote common ancestors.

Gramps is an open source genealogy tool that can produce visualizations based on a family tree specific layout algorithm, it can also display a portrait when available, as well as dates and places of birth and death because they are essential to keep track of who you‘re dealing with, as the data set becomes larger.

The problem was TheBrain graph model is not the same as the one used in Gramps, so Neo4j has been used as a pivot graph technology to convert the data graph structure, with a Python script to operate the transfer.

Visualization issues and solutions, like fine-tuned calculus to make information of interest stand out on a graph display will also be presented (Bloom and yWorks), as well as different interesting uses of graph patterns, that is graph database queries.

Time codes to get to a specific aspect :

00’00 VG & Genealogy usecase introduction
04’30 TheBrain demo : documentation collection, in a graph structure (uncut version of TheBrain demo)
11’30 TheBrain limits
12’50 Gramps visualization and data format
14’58 Neo4j pivot data model
17’05 data transfer process
17’50 generating visualizations with Gramps and Gramps limits
19’10 use of graph topology
21’10 Bloom demo of visual and interactive exploration of the data (stand alone version of Bloom demo)
26’30 yWorks demo to make precisely selected information of interest, stand out on a graph display (stand alone version of yWorks demo)
33’50 wrap up and questions

Nodes 2022 slides

Quoi, où ? mais aussi quand ? une carte avec un curseur de temps (timeline)

Véronique Gendner — Fri, 27 May 2022 18:54:43 +0000

Avoir la bonne clé d’accès à l’information est essentiel pour attraper les informations dont nous avons besoin. Les outils numériques ont l’énorme avantage de permettre des représentations avec plusieurs clés d’accès. La création d’une représentation numérique adaptée passe donc par l’identification des clés d’accès utiles. Voyons par exemple, le cas d’un festival qui présente des spectacles dans différents lieux.

Le site du festival présente les spectacles selon deux clés d’accès :

accès aux informations par spectacle

accès aux informations par lieux

Dans ce cas, une troisième clé d’accès à l’information peut être proposée, c’est l’accès par clé temporelle. Pour cela, le site du festival propose une liste des spectacles pour chaque jour :

Le numérique permet d’aller plus loin dans le croisement de ces trois clés d’accès, tout en gardant une meilleure précision pour la clé temporelle. La solution est de créer une carte avec un curseur de temps:

plein écran

accès aux informations par lieux et par horaire, avec des liens vers les détails des spectacles

L’intérêt est d’avoir sur la même représentation, un accès aux informations par les trois clés, afin de laisser le choix à l’utilisateur, pour qui la clé thématique n’est pas forcement la plus adaptée: lorsqu’un festival présente des spectacles et des Compagnies que l’on ne connaît pas, les clés d’accès par localisation ou horaire peuvent être plus pertinentes.

quels spectacles ont lieu près de chez moi ?
→ localisation sur la carte
- → quel spectacle, avec un lien vers la page comprenant tous les détails
- → quel jour, à quel heure
j’ai du temps dimanche entre 14 et 16h, que puis-je voir ?
→ sélection d’un horaire, sur la carte (à noter que le slider pourrait être amélioré en proposant comme point d’arrêt uniquement les horaires de début d’un spectacle, mais le plugin utilisé ne le permet pas facilement)
- → quel lieux
- → quels spectacles, avec un lien vers la page comprenant tous les détails

Notons également que les clés d’accès à l’information dans l’interface utilisateur ne sont pas nécessairement les mêmes que celles utilisées pour la collecte des données. Dans notre exemple, les informations détaillées étant présentées par spectacle sur le site du festival, il était plus simple de les regrouper ainsi dans le fichier json source pour la carte.

De cette façon, les informations concernant le spectacle (compagnie, style,…) ne sont saisies qu’une fois et associées à une liste de couple (lieu,horaire) pour chaque représentation. C’est le script d’affichage de la carte qui se charge de convertir les données, pour obtenir les clés d’accès voulues dans l’interface utilisateur.

GeoGraph : Normalized Categorization of Multisource Geolocated Data

Véronique Gendner — Mon, 24 Jan 2022 16:51:27 +0000

How GeoGraph, a multisource database of geolocated objects has been build, and how objects categorization is normalized with the Landmarks Objects Ontology build for the Choucas project. We will look into how the graph structure can be used in this context and demo how results of database queries can be displayed on a map, with a QGIS script.

If sound quality is poor, you can watch the video of the presentation here, and for the demo, go back to the above video.

26th Jan 2022 Neo4j Live presentation
This is a more in depth presentation, with a demo, of the brief presentation given at the International Cartographic Conference.

pdf of the presentation

QgisQueryGeoGraph.py : the python script to display GeoGraph in QGIS

GeoGraph Usecase : demo of GASPAR, the UI developed at Laboratoire d’Informatique de Grenoble to help localise the victime, using GeoGraph data (no sound)

Your browser does not support the video tag.

GeoGraph sur le github du projet Choucas

GeoGraph : Modelling and building a graph database of multi-source landmarks to help emergency mountain rescuers

Véronique Gendner — Wed, 15 Dec 2021 14:37:29 +0000

The purpose of the Choucas research project is to come up with methods, tools and resources, to help mountain rescue team localise victims, when answering emergency calls. In this context, we have built a graph data base with the Neo4j Label Property Graph (LPG) technology, that integrates several sources of geolocated objects. Some data comes from the national mapping agency (IGN BDTOPO), others, like routes from crowdsourcing websites. Categorization of multi-source imported data has been normalized by an instantiation process based on the Landmarks Objects Ontology (OOR) that had previously been produced by the project team.
The flexibility of graph databases helps make the right modelling choices by progressively taking into account problems observed in the data as well as researchers and users feedback and new needs.

A more in depth presentation of this work, with a demo of a python script to display the content of the database on a map, with QGIS can be seen here.

Here is the presentation of this work done at the International Cartographic Conference, on 15th Dec. 2021 :

https://www.e-tissage.net/wp-content/uploads/2021/12/T08-2_GendnerVeronique.mp4

T08-2_GendnerVeronique.pdf