Skip to main content
Log in

Type Polymorphism, Natural Language Semantics, and TIL

  • Published:
Journal of Logic, Language and Information Aims and scope Submit manuscript

Abstract

Transparent intensional logic (TIL) is a well-explored type-theoretical framework for semantics of natural language. However, its treatment of polymorphic functions, which are essential for the analysis of various natural language phenomena, is still underdeveloped. In this paper, we address this issue and propose an extension of TIL that introduces polymorphism via type variables ranging over types and generalized variables ranging over constructions and types. Furthermore, we offer an analysis of sentences involving non-specific notional attitudes of the general form ‘A considers (believes, desires, wants, seeks, ...) something’.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. Examples taken from Church (1951, p. 111, 1956, p. 8) and Quine (1956, p. 177).

  2. Similarly to Church (1951), we consider the objects of notional attitudes to be abstract entities rather than mental ones. For an alternative approach, compare, e.g., with Moltmann (2008) and recently Moltmann (2017) utilizing truthmaker semantics.

  3. Fox and Lappin (2005) relies on a similar distinction between schematic and genuine polymorphism, Duží (1993) uses the terms weak and strong polymorphism, and others can no doubt be found as well. It is worth noting that from the perspective of Cardelli and Wegner (1985) the classification (based on Strachey, 2000) of both these kinds of polymorphism would fall under their category of universal/parametric polymorphism. In comparison to Cardelli and Wegner (1985), we rely on a rather strict notion of polymorphism, since they also include in the kinds of polymorphism, e.g., subtyping and overloading.

  4. It is worth noting that, strictly speaking, TIL itself is just an applied instance of Tichý’s type theory (which is a modification of Church’s type theory) intended for the purposes of logical analysis of natural language (similarly to, e.g., transparent hyperintensional logic (THL) recently employed in Raclavský, 2020).

  5. For a proper specification, see Appendix 7. For the definition of n-execution, see Pezlar (2019).

  6. It is important to note that if we allow n-executions with \(n \ge 2\), the Church–Rosser theorem is no longer valid in TIL, as was recently demonstrated by Kosterec (2020).

  7. In a standard TIL presentation, the cases of \(n=0\) (i.e., 0-execution also known as trivialization) and \(n > 0\) are strictly kept apart to emphasize their different conceptual roles, most importantly, 0-execution supplies objects (of any type) for compound constructions, while \((n > 0)\)-execution is used for executing constructions. Furthermore, 0-execution can raise a context of a construction up to the hyperintensional level, while, e.g., 2-execution can decrease the context down (see, e.g., Duží and Horák, 2019). For more about TIL and the three kinds of contexts, see Duží et al. (2010), Sect. 2.6.

  8. TIL is and open-ended framework and other atomic types can be added, e.g., we can add \(\nu \) as the type of natural numbers.

  9. Note that TIL relies on the Church notation \((\alpha \beta )\) for function types. In a more standard notation, this would be written as \(\beta \rightarrow \alpha \). Furthermore, due to the presence of partial functions, we cannot generally assume that all multiargument functions can be reduced to a series of functions taking a single argument. In other words, Schönfinkel’s reduction does not hold. For a proof, see Tichý (1982).

  10. In TIL literature the symbol ‘\(\rightarrow \)’ is used instead of ‘ : ’, however, we choose the latter because it leads to a clearer notation once explicit typing is adopted. Also note that the symbol ‘/’ is used for typing annotations of non-constructions as well. For example, if we want to declare that addition function \(+\) on natural numbers \(\nu \) has type \((\nu \nu \nu )\), we can write it as \(+ / (\nu \nu \nu )\).

  11. It is worth mentioning that Tichý (1988) used explicit subscripts with ‘\(\lambda \)’ to indicate the type of the output of the constructed function. For example, \([\lambda _o x \, [\mathbf {Odd} \; x] ]\).

  12. The need for type variables was also discussed in Raclavský et al. (2015).

  13. TIL-Script, the software variant of TIL, also treats type variables simply as syntactic placeholders and whenever they appear type checking is simply skipped. See e.g., Duží and Fait (2019, p. 223).

  14. An alternative syntactic rule-based approach is, however, also possible. For a brief sketch, see Appendix 7.3.

  15. Parts of the exposition of the variable construction follow my PhD thesis, see Pezlar (2016), pp. 6–8.

  16. We obtain \(v_1{_t}\) from \(v_1\) by replacing objects in the array by their corresponding types, e.g., true is replaced by o, etc.

  17. Note that when we type a variable to a certain type (i.e., specify its range), we are essentially just assigning a concrete number to i. If \(i = 1\), then the variable at hand is typed to construct truth values o, if \(i = 2\), then it constructs natural numbers, etc.

  18. Thus, this T-sequence is essentially obtained by transposing \(v_1{_t}\): from \([[o, o], [\nu , \nu , \nu , \ldots ], [\iota , \iota , \iota , \ldots ], \ldots ]\) we get \([[o, \nu , \iota , \ldots ]]\).

  19. Strictly speaking, we should be distinguishing between valuation and type valuation, but we conflate them to simplify the presentation.

  20. Of course, we could expand valuation arrays to accommodate even them.

  21. An earlier version of this paper contained a more complicated definition of a generalized variable and I would like to thank an anonymous reviewer for suggesting a simplification.

  22. A similar extension of ramified type theory was already briefly discussed in Duží (1993).

  23. Of course, as an anonymous reviewer remarked, the presence of variables is not necessary for forming higher-order objects. For example, the other constructions \([{\textbf {Prime}} \; {{\textbf {3}}}]\) or \([{\textbf {Prime}} \; {{\textbf {4}}}]\) also belong to the second-order type even though they do not contain variables ranging over objects of the first-order type. They belong to the second-order type because they are constructions of order 1 that v-construct objects of a type of order 1 (see Appendix 7, Definition 2). The same holds for constructions belonging to the third-order type, etc. as well.

  24. Note, however, that \(\textsf {Type}\) is not a proper object of eRTT, hence, e.g., a generalized variable x can construct o, but not its type \(\textsf {Type}\).

  25. Recall that ‘\(\exists x : \alpha \ldots \)’ is a notation shortcut for ‘\([\exists \; \lambda x : \alpha \, [\ldots ] ]\)’.

  26. For more about a standard TIL-based analysis of notional attitudes such as seeking, finding, etc., see section 5.2 Notional attitudes in Duží et al. (2010).

  27. This issue was raised by an anonymous reviewer.

  28. Examples taken from Duží et al. (2010).

  29. Note that in contrast to, e.g., second-order lambda calculus (Girard 1972; Reynolds 1974), the logic of construction-terms and type-terms remains separated.

References

  • Cardelli, L., & Wegner, P. (1985). On understanding types, data abstraction, and polymorphism. ACM Computing Surveys, 17(4), 471–523. https://doi.org/10.1145/6041.6042

    Article  Google Scholar 

  • Chierchia, G. (1982). Nominalization and Montague grammar: A semantics without types for natural languages. Linguistics and Philosophy, 5(3), 303–354. https://doi.org/10.1007/BF00351458

    Article  Google Scholar 

  • Church, A. (1951). The need for abstract entities in semantic analysis. Proceedings of the American Academy of Arts and Sciences, 80(1), 100–112.

    Article  Google Scholar 

  • Church, A. (1956). Introduction to mathematical logic. Princeton University Press.

  • Duží, M. (1993). Frege, notional attitudes, and the problem of polymorphism. In M. Stelzner & W. Stelzner (Eds.), Logik und mathematik Frege-Kolloquium Jena (pp. 314–323). de Gruyter.

  • Duží, M., & Fait, M. (2019). Type checking algorithm for the TIL-Script language. In T. Endrjukaite, A. Dudko, H. Jaakkola, B. Thalheim, Y. Kiyoki, & N. Yoshida (Eds.), Information modelling and knowledge bases XXX, frontiers edn. IOS Press. https://doi.org/10.3233/978-1-61499-933-1-219

  • Duží, M., & Horák, A. (2019). Hyperintensional reasoning based on natural language knowledge base. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems. http://arxiv.org/abs/1906.07562.

  • Duží, M., Jespersen, B., & Materna, P. (2010). Procedural semantics for hyperintensional logic: Foundations and applications of Transparent Intensional Logic. Springer. https://doi.org/10.1007/978-90-481-8812-3

  • Fox, C., & Lappin, S. (2005). Foundations of intensional semantics. Blackwell.

  • Girard, J. Y. (1972). Interprétation fonctionnelle et Élimination des coupure de l’arithmétique d’ordre supérieur. Ph.D thesis, Université Paris VII.

  • Kosterec, M. (2020). Substitution contradiction, its resolution and the Church–Rosser Theorem in TIL. Journal of Philosophical Logic, 49(1), 121–133. https://doi.org/10.1007/s10992-019-09514-y

    Article  Google Scholar 

  • Moltmann, F. (2008). Intensional verbs and their intentional objects. Natural Language Semantics, 16(3), 239–270. https://doi.org/10.1007/s11050-008-9031-5

    Article  Google Scholar 

  • Moltmann, F. (2017). Cognitive products and the semantics of attitude verbs and deontic modals. In F. Moltmann & M. Textor (Eds.), Act-based conceptions of propositional content. (Vol. 408). Oxford University Press.

  • Pezlar, I. (2016). Investigations into Transparent Intensional Logic: A rule-based approach. Ph.D thesis, Masaryk University. https://is.muni.cz/th/hhhga/pezlar_phd_thesis.pdf

  • Pezlar, I. (2017). Algorithmic theories of problems. A constructive and a non-constructive approach. Logic and Logical Philosophy, 26(4), 473–508. https://doi.org/10.12775/LLP.2017.010

    Article  Google Scholar 

  • Pezlar, I. (2019). On two notions of computation in Transparent Intensional Logic. Axiomathes. https://doi.org/10.1007/s10516-018-9401-7

  • Quine, W. V. O. (1956). Quantifiers and propositional attitudes. Journal of Philosophy, 53(5), 177–187. https://doi.org/10.2307/2022451

    Article  Google Scholar 

  • Raclavský, J. (2020). Belief attitudes, fine-grained hyperintensionality and type-theoretic logic. College Publications.

  • Raclavský, J., Kuchyňka, P., & Pezlar, I. (2015). Transparentní intenzionální logika jako characteristica universalis a calculus ratiocinator. Masaryk University Press (Munipress).

  • Reynolds, J. C. (1974). Towards a theory of type structure. In Colloquium on programming, Paris, 9–11 April 1974 (pp. 1–18).

  • Strachey, C. (2000). Fundamental concepts in programming languages. Higher-Order and Symbolic Computation, 13(1/2), 11–49. https://doi.org/10.1023/A:1010000313106

    Article  Google Scholar 

  • Tichý, P. (1982). Foundations of partial type theory. Reports on Mathematical Logic, 14, 59–72. https://doi.org/10.1007/BF00370346

    Article  Google Scholar 

  • Tichý, P. (1988). The foundations of Frege’s logic. de Gruyter.

Download references

Acknowledgements

I would like to thank Prof. Marie Duží and Prof. Jiří Raclavský for their valuable comments that helped to significantly improve this paper.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Ivo Pezlar.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Work on this paper was supported by Grant No. 19-12420S from the Czech Science Foundation, GA ČR.

Appendix

Appendix

1.1 Constructions

The original definition of TIL constructions was given by Tichý (1988), here we follow Duží et al. (2010) (an alternative formulation can be found in Raclavský et al., 2015):

Definition 7

(Constructions)

  1. 1.

    The variable x is a construction that constructs an object O of the respective type dependently on a valuation. We say that it v-constructs O.

  2. 2.

    Where X is any object, \(^0 X\) is the construction trivialization. It constructs X without any change.

  3. 3.

    The composition \([X_0 \, X_1 \ldots X_m]\) is the following construction. If X v-constructs a function f of type \((\alpha \beta _1 \ldots \beta _m)\) and \(X_1 \ldots X_m\) v-construct objects \(b_1 , \ldots , b_m\) of types \(\beta _1 , \ldots , \beta _m\), respectively, then the composition \([X_0 \, X_1 \ldots X_m]\) v-constructs the value (an object of type \(\alpha \), if any) of f on the tuple-argument \(\langle b_1 , \ldots ,b_m \rangle \). Otherwise, it is a v-improper construction, i.e., construction that does not construct anything.

  4. 4.

    The closure \([\lambda x_1 \ldots x_m \, Y]\) is the following construction. Let \(x_1 , \ldots x_m\) be pairwise distinct variables v-constructing objects of types \(\beta _1 , \ldots , \beta _m\) and Y a construction v-constructing an object of type \(\alpha \). Then \([\lambda x_1 \ldots x_m \, Y]\) is the construction closure. It v-constructs the following function f of type \((\alpha \beta _1 \ldots \beta _m)\): let \(\langle b_1, \ldots , b_m \rangle \) be a tuple of objects of types \(\beta _1 \ldots \beta _m\), respectively, and \(v'\) be a valuation that associates \(x_i\) with \(b_i\) and is identical to v otherwise. Then the value of function f on argument tuple \(\langle b_1, \ldots , b_m \rangle \) is the object of type \(\alpha \) \(v'\)-constructed by Y. If Y is \(v'\)-improper, then f is undefined on \(\langle b_1, \ldots , b_m \rangle \).

  5. 5.

    The single execution \(^1 X\) is the construction that either v-constructs the object v-constructed by X or, if X v-constructs nothing, is v-improper.

  6. 6.

    The double execution \(^2 X\) is the following construction: let X be any object, the double execution \(^2 X\) is v-improper if X is a non-construction or if X does not v-construct a construction or if X v-constructs a v-improper construction. Otherwise, let X v-construct a construction \(X'\) and let \(X'\) v-construct and object \(X''\), then \(^2 K\) v-constructs \(X''\).

  7. 7.

    Nothing other is a construction, unless it follows from 1 to 6.

1.2 Ramified Type Theory

We follow the specification from Tichý (1988):

Definition 8

(Ramified type theory of TIL)

Let B be a base, i.e., a set of atomic types.

  1. 1.
    (\(\hbox {t}_1\)i):

    Every member of B is a type of order 1 over B.

    (\(\hbox {t}_1\)ii):

    If \(0<m\) and \(\alpha , \beta _1 , \ldots , \beta _m\) are types of order 1 over B, then the collection \((\alpha \beta _1 \ldots \beta _m)\) of all m-ary (total and partial) mappings from \(\beta _1 ,\ldots , \beta _m\) to \(\alpha \) is also a type of order 1 over B.

    (\(\hbox {t}_1\)iii):

    Nothing is a type of order 1 over B unless it follows from (\(\hbox {t}_1\)i) and (\(\hbox {t}_1\)ii).

  2. 2.
    (\(\hbox {c}_k\)i):

    Let \(\alpha \) be any type of order k over B. Every variable ranging over \(\alpha \) is a construction of order k over B. If X is of (i.e., belongs to) type \(\alpha \), then \(^0 X\), \(^1 X\), and \(^2 X\) are constructions of order k over B.

    (\(\hbox {c}_k\)ii):

    If \(0<m\) and \(X_0, X_1, \ldots , X_m\) are constructions of order k, then \([X_0 \; X_1 \ldots \, X_m]\) is a construction of order k over B. If \(0<m\), \(\alpha \) is a type of order k over B, and Y as well as the distinct variables \(x_1, \ldots , x_m\) are constructions of order k over B, then \([\lambda _\alpha \, x_1 \ldots x_m \; Y]\) is a construction of order k over B.

    (\(\hbox {c}_k\)iii):

    Nothing is a construction of order k over B unless it follows from (\(\hbox {c}_k\)i) and (\(\hbox {c}_k\)ii).

Let \(*_k\) be the collection of constructions of order k over B. The collection of types of order \(k+1\) over B is defined as follows:

(\(\hbox {t}_{k+1}\)i):

\(*_k\) and every type of order k is a type of order \(k+1\).

(\(\hbox {t}_{k+1}\)ii):

If \(0<m\) and \(\alpha , \beta _1 , \ldots , \beta _m\) are types of order \(k+1\) over B, then the collection \((\alpha \beta _1 \ldots \beta _m)\) of all m-ary (total and partial) mappings from \(\beta _1 , \ldots , \beta _m\) to \(\alpha \) is also a type of order \(k+1\) over B.

(\(\hbox {t}_{k+1}\)iii):

Nothing is a type of order \(k+1\) over B unless it follows from (\(\hbox {t}_{k+1}\)i) and (\(\hbox {t}_{k+1}\)ii).

1.3 Types and Rules

In this paper, we have explored a semantic treatment of type variables in TIL. Alternatively, we could attempt a syntactic rule-based analysis as well. However, since this topic is outside the scope of the present paper, we sketch only the basics of this approach. The key observation is that type variables generally appear in two kinds of judgments in TIL literature:

$$\begin{aligned}&C : *_n \rightarrow \sigma (\alpha )\\&X : \sigma (\alpha ) \end{aligned}$$

which can be read as ‘a construction C belonging to a type \(*_n\) is typed to v-construct an object of type \(\sigma \)’ and ‘a non-construction X belongs to a type \(\alpha \)’, respectively, where \(\sigma \) may contain \(\alpha \) as a free type variable. For exampleFootnote 28:

\(Cardinality_\tau : (\tau (o \tau ))\)

\(Cardinality_\alpha : (\tau (o \alpha ))\)

\(Tr_\alpha : (*_n \alpha ) \)

\(X : *_n \rightarrow ((\alpha \tau ) \omega )\)

\(z : *_1 \rightarrow \alpha \)

Furthermore, note that these judgments are composed of three parts: a typed object (i.e., either a construction C or a non-construction X), a typing relation ‘ : ’ and a type term (i.e., either \(*_n \rightarrow \sigma (\alpha )\) in case of constructions or \(\sigma (\alpha )\) in case of non-constructions).

Now, as we mentioned above, type terms might depend on some variable \(\alpha \) of type \(\textsf {Type}\) (the type of all types). We can explicitly capture this dependency by abstracting type terms from \(\alpha \) via \({\uplambda }\) abstractor and get \({\uplambda }\)\(\alpha : \textsf {Type} . *_n \rightarrow \sigma (\alpha )\) and \({\uplambda }\)\(\alpha : \textsf {Type} . \sigma (\alpha )\). Consequently, every free occurrence of \(\alpha \) in \(\sigma \) becomes bound in \({\uplambda }\)\(\alpha : \textsf {Type} . *_n \rightarrow \sigma (\alpha )\) and \({\uplambda }\)\(\alpha : \textsf {Type} . \sigma (\alpha )\). The obvious complement to abstraction is, of course, application. Thus, the rules we obtain are as follows:Footnote 29

figure a

where \(\sigma [\kappa /\alpha ]\) is the result of substituting \(\kappa \) for all occurrences of \(\alpha \) in \(\sigma \).

Note that the premises of abs-C and abs-X rules are hypothetical judgments, i.e., judgments made in a certain context. Thus, we can read \( \alpha : \mathsf {Type} \vdash C : *_n \rightarrow \sigma (\alpha )\) as ‘a judgment \(C : *_n \rightarrow \sigma (\alpha )\) is assertable given that we have some type term \(\alpha \) of type \(\textsf {Type}\)’. Analogously for the hypothetical judgment \( \alpha : \mathsf {Type} \vdash X : \sigma (\alpha )\).

These rules can help us to explain syntactically the general process of instantiation of type terms containing type variables to some specific type, a process which was investigated semantically in this paper via the use of expanded valuation arrays.

Rights and permissions

Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Pezlar, I. Type Polymorphism, Natural Language Semantics, and TIL. J of Log Lang and Inf 32, 275–295 (2023). https://doi.org/10.1007/s10849-022-09383-w

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10849-022-09383-w

Keywords

Navigation