FSU-HuME Joint Conference
Relations and Elements
November 20, 2026 – Milano, Italy
Neural Language Models and the Renewal of the Structuralist Hypothesis
Il n’y a pas de “philosophie” du langage.
Il n’y a que la linguistique.
Louis Hjelmslev, Principes de grammaire générale (1928)
Hjelmslev (1961)
“A totality does not consist of things but of relationships”
Lévi-Strauss (1964)
“Two syntagmatic chains, or fragments of the same chain, which, taken separately, offered no definite meaning, acquire one from the mere fact that they are opposed”
Ducrot (1968)
“Either one starts from the system, or one never finds the elements”
Sperber (1968)
“It is the relations between the elements, and not the elements themselves, that define the structure”
Wahl (1968)
“There are not first signs, but always first languages, that is to say a structure”
“The elements of a structure have neither extrinsic designation nor intrinsic signification. (…) They have nothing but a sense: a sense which is necessarily and uniquely one of ‘position’.”
“Places in a purely structural space are primary in relation to the things and real beings that come to occupy them.”
Deleuze (2002, p. 243)
“It is right to assign linguistics as the origin of structuralism: not only Saussure, but the Moscow school, the Prague school.”
Deleuze (2002, p. 239)
“It is a great illusion to consider a term simply as the union of a certain sound with a certain concept. (…) it would be to believe that one can start from the terms and build the system by adding them up, whereas, on the contrary, one must start from the interdependent whole to obtain by analysis the elements it contains.” (157)
weekend – weak end
ice cream – I scream
new direction – nude erection
Poor old Dali loped with an amazin’ raging cyst,
as poor Roald Dahl eloped with Anna-May’s enraging sisters.
les dents, la bouche
les dents la bouchent
l’aidant la bouche
laides en la bouche
lait dans la bouche
…
“Language then has the strange, striking characteristic of not having entities that are perceptible at the outset and yet of not permitting us to doubt that they exist and that their functioning constitutes it. Doubtless we have here a trait that distinguishes language from all other semiological institutions.”
Saussure (1959, p. 107)



“I call the combination of a concept and a sound image a sign…”
“I propose to retain the word sign to designate the whole and to replace concept and sound-image respectively by signified [signifié] and signifier [signifiant]”


“We might consider assigning elements to the same category only if they have all contexts in common (i.e., are mutually substitutable in all sentences), if they share some context, or if the ancestral of this latter relation holds between them.”
Chomsky (1953)


Chomsky (1953)

Chomsky (1953)

Chomsky (1953)

| S1: | (a, c), (d, f) |
| S2: | (a…d, c…f) |
| S3: | (x, y) |
The preceding discussion probably does not exhaust the constructional possibilities within the limits of this general program, but until a more refined set of criteria of adequacy is given, and along with it, a refined analysis of kinds of language, it seems premature to enter into an exhaustive analysis of all possible systems of this general form.
Chomsky (1953, p. 256)



Chomsky (1957, p. 27)
Fitch (2014)

Chomsky (1995, p. 19)
“For each particular language, the cognitive system, we assume, consists of a computational system CS and a lexicon. The lexicon specifies the elements that CS selects and integrates to form linguistic expressions—(PF, LF) pairings, we assume. The lexicon should provide just the information that is required for CS, without redundancy and in some optimal form, excluding whatever is predictable by principles of UG or properties of the language in question. Virtually all items of the lexicon belong to the substantive categories, which we will take to be noun, verb, adjective, and particle, putting aside many serious questions about their nature and interrelations.”
Chomsky (1995, p. 6)



Firth (1957)
“You shall know a word by the company it keeps!”
Rubenstein & Goodenough (1965)
“Words which are similar in meaning occur in similar contexts”
Schütze & Pedersen (1995)
“Words with similar meanings will occur with similar neighbors if enough text material is available”
Landauer & Dumais (1997)
“A representation that captures much of how words are used in natural context will capture much of what we mean by meaning”
Pantel (2005)
“Words that occur in the same contexts tend to have similar meanings”
Lenci (2008)
“The degree of semantic similarity between two linguistic expressions A and B is a function of the similarity of the linguistic contexts in which A and B can appear”
\(\ell = \sum\limits_{w\in V_w} \sum\limits_{c\in V_c} \#(w,c) \left(\log \sigma (\vec{w} \cdot \vec{c} ) + k \cdot \mathbb{E}_{c_N \sim P_D} [\log \sigma(-\vec{w}\cdot \vec{c}_N)]\right)\)
\(\dfrac{\partial \ell }{\partial (\vec{w} \cdot \vec{c})} = 0 \quad\text{when}\quad \vec{w} \cdot \vec{c} = \log \left(\dfrac{\#(w,c) \cdot |D|}{\#(w)\cdot \#(c)}\right) - \log k\)
\(\vec{w} \cdot \vec{c} = {\color{#ff3333}{\mathrm{PMI}(w,c)}} - \log k\)
word2vec performs an implicit, low-dimensional factorization of a pointwise mutual information word-context matrix
The Singular Value Decomposition (SVD) provides an exact solution to this problem
\(\color{red}{W} = \{\texttt{-},\texttt{/},\texttt{0},\texttt{1},\texttt{2},\texttt{3},\texttt{4},\texttt{5},\texttt{6},\texttt{7},\texttt{8},\texttt{9},\texttt{=},\texttt{a},\texttt{b},\texttt{c},\ldots,\texttt{w},\texttt{x},\texttt{y},\texttt{z},\texttt{é}\}\)
\(\color{blue}{C} = W \times W = \{\texttt{(-,-)},\texttt{(-,/)},\texttt{(-,0)},\ldots,\texttt{(é,z)},\texttt{(é,é)}\}\)
\(\begin{aligned} M_{\color{red}{w}\color{blue}{c}} &= \mathrm{pmi}(\color{red}{w},\color{blue}{c})\\ &= \log \frac{p(\color{red}{w},\color{blue}{c})}{p(\color{red}{w})\,p(\color{blue}{c})} \end{aligned}\)
\(U\)

\(\Sigma\)

\(V^T\)

\(U \times \Sigma\)
\(\hat{U} \times \hat{\Sigma}\)

\(\hat{U} \times \hat{\Sigma}\)


| your | : | house |
| my | : | apartment |
“A priori it would seem to be a generally valid thesis that for every process there is a corresponding system, by which the process can be analyzed and described by means of a limited number of premises. It must be assumed that any process can be analyzed into a limited number of elements recurring in various combinations. Then, on the basis of this analysis, it should be possible to order these elements into classes according to their possibilities of combination. And it should be further possible to set up a general and exhaustive calculus of the possible combinations.”
Hjelmslev (1953, p. 9)
\(\color{red}{W} = \{\texttt{-},\texttt{/},\texttt{0},\texttt{1},\texttt{2},\texttt{3},\texttt{4},\texttt{5},\texttt{6},\texttt{7},\texttt{8},\texttt{9},\texttt{=},\texttt{a},\texttt{b},\texttt{c},\ldots,\texttt{w},\texttt{x},\texttt{y},\texttt{z},\texttt{é}\}\)
\(\color{blue}{C} = W \times W = \{\texttt{(-,-)},\texttt{(-,/)},\texttt{(-,0)},\ldots,\texttt{(é,z)},\texttt{(é,é)}\}\)
\[\begin{aligned} M \colon {\color{red}{W}} \times {\color{blue}{C}} &\to \mathbb{R}\\ ({\color{red}{w}},{\color{blue}{c}}) &\mapsto \mathrm{pmi}({\color{red}{w}},{\color{blue}{c}}) \end{aligned}\]
\[\begin{aligned} M_w \colon {\color{red}{W}} &\to \mathbb{R}^{\color{blue}{C}}\\ {\color{red}{w}} &\mapsto {\color{blue}{M(w,-)}} \end{aligned}\]
\[\begin{aligned} M_c \colon {\color{blue}{C}} &\to \mathbb{R}^{\color{red}{W}}\\ {\color{blue}{c}} &\mapsto {\color{red}{M(-,c)}} \end{aligned}\]
\(M^* \colon \mathbb{R}^{\color{red}{W}} \to \mathbb{R}^{\color{blue}{C}}\)
\(M_* \colon \mathbb{R}^{\color{blue}{C}} \to \mathbb{R}^{\color{red}{W}}\)
\({\color{red}{M_* M^*}} \colon \mathbb{R}^{\color{red}{W}} \to \mathbb{R}^{\color{red}{W}}\)
\({\color{blue}{M^* M_*}} \colon \mathbb{R}^{\color{blue}{C}} \to \mathbb{R}^{\color{blue}{C}}\)
\[\begin{aligned} \{{\color{red}{u_1}}, \ldots, {\color{red}{u_m}}\} &\subset \mathbb{R}^{\color{red}{W}}\\ \{{\color{blue}{v_1}}, \ldots, {\color{blue}{v_n}}\} &\subset \mathbb{R}^{\color{blue}{C}}\\ \{\lambda_1, \ldots, \lambda_{\min(m,n)}&, 0, \ldots, 0\} \end{aligned}\]
\({\color{red}{M_* M^* u_i}} = \lambda_i {\color{red}{u_i}}\)
\({\color{blue}{M^* M_* v_i}} = \lambda_i {\color{blue}{v_i}}\)
The \({\color{red}{u_i}}\) and \({\color{blue}{v_i}}\) are (linear)
fixed points!
\[M = U\Sigma V^T \qquad \begin{aligned} U &:= [{\color{red}{u_1}}, \ldots, {\color{red}{u_m}}]\\ V &:= [{\color{blue}{v_1}}, \ldots, {\color{blue}{v_n}}]\\ \Sigma &:= \left[\begin{smallmatrix} \sqrt{\lambda_1} & \cdots & 0 \\ \vdots & \ddots & \vdots \\ 0 & \cdots & \sqrt{\lambda_r} \end{smallmatrix}\right] \end{aligned}\]

\(M^*\)

\(M_*\)

\({\color{red}{M_* M^*}}\)
\({\color{red}{M_* M^* u}} = \lambda {\color{red}{u}}\)
\(\times\)
\(=\) \(\begin{bmatrix} \lambda_1 & \lambda_2 & \lambda_3 \end{bmatrix}\) 

(credits: Jacopo Bertolotti, CC0
via Wikimedia Commons)



fixed point!
Eigenvectors of \({\color{red}{M_* M^*}}\):

Eigenvalues of \({\color{red}{M_* M^*}}\) and \({\color{blue}{M^* M_*}}\):

Eigenvectors of \({\color{blue}{M^* M_*}}\):


\(M \colon {\color{red}{W}} \times {\color{blue}{C}} \to \mathbb{R}\)
\({\color{orange}{\mathsf{C}}}^{\mathrm{op}} \times {\color{green}{\mathsf{D}}} \to \mathsf{Set}\)
\[\left.\begin{array}{l} {\color{red}{M_* M^*}} \colon \mathbb{R}^{\color{red}{W}} \to \mathbb{R}^{\color{red}{W}} \\ {\color{blue}{M^* M_*}} \colon \mathbb{R}^{\color{blue}{C}} \to \mathbb{R}^{\color{blue}{C}} \end{array}\right\}\]
distributional
operator
\[\left\{\begin{array}{l} {\color{orange}{\mathcal{M}_* \mathcal{M}^*}} \colon \mathsf{Set}^{{\color{orange}{\mathsf{C}}}^{\mathrm{op}}} \to \mathsf{Set}^{{\color{orange}{\mathsf{C}}}^{\mathrm{op}}} \\ {\color{green}{\mathcal{M}^* \mathcal{M}_*}} \colon (\mathsf{Set}^{\color{green}{\mathsf{D}}})^{\mathrm{op}} \to (\mathsf{Set}^{\color{green}{\mathsf{D}}})^{\mathrm{op}} \end{array}\right.\]
\[\left.\begin{array}{l} \{{\color{red}{u_i}} \in \mathbb{R}^{\color{red}{W}} \mid {\color{red}{M_* M^*}}\, {\color{red}{u_i}} = \lambda_i {\color{red}{u_i}}\} \\ \{{\color{blue}{v_i}} \in \mathbb{R}^{\color{blue}{C}} \mid {\color{blue}{M^* M_*}}\, {\color{blue}{v_i}} = \lambda_i {\color{blue}{v_i}}\} \end{array}\right\}\]
fixed
points
\[\left\{\begin{array}{l} \{{\color{orange}{f}} \in \mathsf{Set}^{{\color{orange}{\mathsf{C}}}^{\mathrm{op}}} \mid {\color{orange}{\mathcal{M}_* \mathcal{M}^*}}({\color{orange}{f}}) \cong {\color{orange}{f}}\} \\ \{{\color{green}{g}} \in (\mathsf{Set}^{\color{green}{\mathsf{D}}})^{\mathrm{op}} \mid {\color{green}{\mathcal{M}^* \mathcal{M}_*}}({\color{green}{g}}) \cong {\color{green}{g}}\} \end{array}\right.\]
\({\color{red}{M_* M^* u}} = \lambda {\color{red}{u}}\)
\(\times\)
\(=\) \(\begin{bmatrix} \lambda_1 & \lambda_2 & \lambda_3 \end{bmatrix}\) 
\({\color{orange}{\mathcal{M}_* \mathcal{M}^*}}\, {\color{orange}{f}} = {\color{orange}{f}}\)
\(\star\)
\({\color{#ff3333}{\stackrel{?}{=}}}\) 
\({\color{orange}{\mathcal{M}_* \mathcal{M}^*}}\, {\color{orange}{f}} = {\color{orange}{f}}\)











fixed
points!!!
fixed
points!!!
“If it is true […] that, first, the dominant class constitutes a relatively autonomous space whose structure is defined by the distribution of economic and cultural capital among its members, […] then these structures should be found in the space of life-styles […]. To endeavour to establish this, the whole set of survey data was subjected to correspondence analysis.”
Gastaldi, J. L. (2021). Why Can Computers Understand Natural Language? Philosophy & Technology, 34(1), 149–214. https://doi.org/10.1007/s13347-020-00393-9
Gastaldi (2021)
Gastaldi, J. L., & Pellissier, L. (2021). The Calculus of Language: Explicit Representation of Emergent Linguistic Structure through Type-Theoretical Paradigms. Interdisciplinary Science Reviews, 46(4), 569–590. https://doi.org/10.1080/03080188.2021.1890484
Gastaldi & Pellissier (2021)
Bradley, T.-D., Gastaldi, J. L., & Terilla, J. (2024). The Structure of Meaning in Language: Parallel Narratives in Linear Algebra and Category Theory. Notices of the American Mathematical Society.
Bradley et al. (2024)
Gastaldi, J. L., Jarvis, S., Seiller, T., & Terilla, J. (2026). Projective metric geometry of tropical nuclei: Gap matrices, event loci, and order chambers. https://arxiv.org/abs/2601.07900
Gastaldi et al. (2026b)
Gastaldi, J. L., Jarvis, S., Seiller, T., & Terilla, J. (2026). A calculus of types in isbell nuclei. https://arxiv.org/abs/2606.03369
Gastaldi et al. (2026a)
Gianni Gastaldi | The Structure Before the Tokens